Contents

D. Intrinsic CriticismE. Critical Freedom and Interpretive Constraint

Chapter 5 PROBLEMS AND PRINCIPLES OF VALIDATION

A. The Self-Confirmability of InterpretationsB. The Survival of the FittestC. The Logic of Validation: Principles of ProbabilityD. The Logic of Validation: Interpretive EvidenceE. Methods, Canons, Rules, and Principles

Appendix I OBJECTIVE INTERPRETATION

A. The Two Horizons of Textual MeaningB. Determinateness of Textual MeaningC. Verification

Appendix II GADAMER'S THEORY OF INTERPRETATION

A. Tradition and the Indeterminacy of MeaningB. Repetition and the Problem of NormsC. Explication and the Fusion of HorizonsD. The Historicity of UnderstandingE. Prejudice and Pre-Understanding

Appendix III AN EXCURSUS ON TYPES

A. Self-Identity of TypesB. Verbal Meanings as Types

Index

f ,\). t{; h( [;~

\lvt ~ ,,, I)\h rM~\;Q'"

xiv

144155

164164169173180198

209212224235

245247251252254258

265265269

275

1.iN DEFENSE OF THE AUTHOR

It has been said of Boehme that his books arelike a picnic to which the author brings the wordsand the reader the meaning. The remark mayhave been intended as a sneer at Boehme, but itis an exact description of all works of literary artwithout exception.

Northrop Frye

,{ A,' BANISHMENT OF THE AUTHOR

j''1.~i; a task for the historian of culture to explain why there haspeen in the past four decades a heavy and largely victorious~ss.i:m1t on the sensible belief that a text means what its authornIeant. In the earliest and most decisive wave of the attack(launched by Eliot, Pound, and their associates) the bat,:ground was literary: the proposition that textual meaning i;,'independent of the author's control was associated with t11c

.,)it~~ary doctrine that the best poetry is impersonal, objective,':·';J~ild. autonomous; that it leads an afterlife of its own, totally;,;:;?,:;~~t off from the life of its author. 1 This programmatic notion'",vhat poetry should be became subtly identified with a notion

,. at what all poetry and indeed all forms of literature necessarily>W-h~tbe. It was not simply desirable that literature should de-"s[.!~~li itself from the ~ubjective realm of the author's personal

fhoughts and feelings; it was, rather, an indubitable fact that: <;ll~ written language remains independent of that subjective',;.i:f~'aJm. At a slightly later period, and for different reasons, this

•~ame notion of semantic autonomy was advanced by Heidegger

'J3:i~! 1. The classic statement is in T. S. Eliot, "Tradition and the Indi­., vidual Talent," Selected Essays (New York, 1932).

1

/,

t·~'------"""'----------------- I:

Chapter 1': In Defense of the Author

and his followers. 2 The idea also has been advocated by writerswho believe with Jung that individual expressions may quiteunwittingly express archetypal, communal meanings. In somebranches of linguistics, particularly in so-called informationtheory, the semantic autonomy of language has been a workingassumption. The theory has found another home in the work ofnon-Jungians who have interested themselves (as Eliot didearlier) in symbolism, though Cassirer, whose name is some­times invoked by such writers, did not believe in the semanticautonomy of language. 3 As I said, it is the job of the culturalhistorian to explain why this doctrine should have gained cur­rency in recent times, but it is the theorist's job to determinehow far the theory of semantic autonomy deserves acceptance.

Literary scholars have often contended that the theory ofauthorial irrelevance was entirely beneficial to literary criticismand scholarship because it shifted the focus of discussion fromthe author to his work. Made confident by the theory, themodern critic has faithfully and closely examined the text toferret out its independent meaning instead of its supposedsignificance to the author's life. That this shift toward exegesishas been desirable most critics would agree, whether or notthey adhere to the theory of semantic autonomy. But the theoryaccompanied the exegetical movement for historical not logicalreasons, since no logical necessity compels a critic to banish anauthor in order to analyze his text. Nevertheless, through itshistorical association with close exegesis, the theory has lib­erated much subtlety and intelligence. Unfortunately, it hasalso frequently encouraged willful arbitrariness and extrava­gance in academic criticism and has been one very importantcause of the prevailing skepticism which calls into doubt thepossibility of objectively valid interpretation. These disadvan-

2. See, for example, Martin Heidegger, Unterwegs zur Sprache(Pfullingen, 1959).

3. See Ernst Cassirer, The Philosophy of Symbolic Forms: Vol. 1,Language, trans. R. Manheim (New Haven, 1953), particularly pp. 69,178, 213, 249-50, and passim.

2

A. Banishment of the Author

tages would be tolerable, of course, if the theory were true. Inintellectual affairs skepticism is preferable to illusion.

The disadvantages of the theory could not have been easilypredicted in the exciting days when the old order of academiccriticism was being overthrown. At that time such naIvetes asthe positivistic biases of literary history, the casting about forinfluences and other causal patterns, and the post-romanticfascination with the habits, feelings, and experiences surround­ing the act of composition were very justly brought under at­tack. It became increasingly obvious that the theoretical foun­dations of the. old criticism were weak and inadequate. Itcannot be said, therefore, that the theory of authorial irrele­vance was inferior to the theoxies or quasi-theories it replaced,nor can it be doubted that the immediate effect of banishingthe author was wholly beneficial and invigorating. Now, at adistance of several decades, the difficulties that attend thetheory of semantic autonomy have clearly emerged and areresponsible for that uneasiness which persists in the academies,although the theory has long been victorious.

That this state ofacaclemic skepticism and disarray resultslargely from the theory ,of authorial irrelevance is, J think, a factof our recent intellectual history. For, once the author had beenruthlessly banished as the determiner of his text's meaning, itvery gradually appeared that no adequate principle existed forjudging the validity of an interpretation. By an inner necessitythe study of "what a text says" became the study of what it saysto an individual critic. It became fashionable to talk about acritic's "reading" of a text, and this word began to appear inthe titles of scholarly works. The word seemed to imply thatif the author had been banished, the critic still remained, andhis new, original, urbane, ingenious, or relevant "reading"

[

carried its own interest.What had not been noticed in the earliest enthusiasm for

going back to "what the te;xt says" was that the text had torepresent somebody's meaning-if not the author's, then the

:. critic's. It is true that a theory was erected under which the

·1

Ii

II.I:

rl-II

Chapter 1: In Defense of the Author

meaning of the text was equated with everything it couldplausibly be taken to mean. (I have described in Appendix Ithe fallacies of this and other descriptions of meaning that were

. contrived to escape the difficulties of authorial irrelevance. 4 )

The theory of semantic autonomy forced itself iq.to such un­satisfactory, ad hoc formulations because in its zeal to banishthe author it ignored the fact that meaning is an affair of con-

.+ sciousness not of words. Almost any word sequence can, underthe conventions of language, legitimately represent more thanone complex of meaning. 5 A word sequence means nothing inparticular until somebody either means something by it orunderstands something from it. There is no magic land ofmeanings outside human consciousness. Whenever meaning isconnected to words, a person is making the connection, andthe particular meanings he lends to them .are never the onlylegitimate ones under the .norms and conventions of his lan­guage.

One proof that the conventions of language can sponsordifferent meanings from the same sequence of words resides inthe fact that interpreters can and do disagree. When these dis­agreements occur, how are they to be resolved? Under thetheory of semantic autonomy they cannot be resolved, sincethe meaning is not what the author meant, but "what the poemmeans to different sensitive readers."G One interpretation is asvalid as another, so long as it is "sensitive" or "plausible." Yetthe teacher of literature who adheres to Eliot's theory is also byprofession the preserver of a heritage and the conveyor ofknowledge. On what ground does he cI"aim that his "reading"is more valid than that of any pupil? On no very firm ground.

4. See particularly pp. 224-35.5. The random example that I use later in the book is the sentence:

"1 am going to town today." Different senses can be lent to the sentenceby the simple device of placing a strong emphasis on any of the sixdifferent words.

6. The phrase is from T. S. Eliot, On Poetry and Poets (New York,1957), p. 126.

4

A. Banishment of the Author

~his impasse is a principal cause of the loss of bearings senne­tImes felt though not often confessed by academic critics.

One ad hoc theory that has been advanced to circumvenl L'j'_chaotic democracy of "readings" deserves special mentio1111cT.'because it involves the problem of value, a problem that p~_>

occupies some modern literary theorists. The most valid Ti~'Ic1..

ing of a text is the "best" reading. 7 But even if we assumed thata critic did have access to the divine criteria by which he cou1;:;determine the best reading, he would still be left with twnequally compelling normative ideals-the best meaning aJ1cl

the author's meaning. Moreover, if the best meaning were j!U~

the author's, then it would have to be the critic's-in Wh>Jlcase the critic would be the author of the best meaning. W;,'~ll­

ever meaning is attached to a sequence of words it is impossj;:lbto escape an author.

Thus, when critics deliberately banished the original author,they themselves usurped his place, and this led unerringly tosome of our present-day theoretical confusions. Where b~j:orethere had been but one author, there now arose a multiplicityof tbem, each carrying as much authority as the next. To b~li' ;;.,the original author as the determiner of meaning was to IT ' ..>.;i

the only compelling normative principle that could lend validiLyto an interpretation. On the other hand, it might be the casethat there does not really exist a viable normative ideal thatgoverns the interpretation of texts. This would follow if any ofthe various arguments brought against the author were to hold.For if the meaning of a text is not the author's, then no inter-

~ pretation can possibly correspond to the meaning of the text,since the text can have no determinate or determinable mean-

7. It would be invidious to name any individual critic as the be­getter of this widespread and imprecise I?otion. By the "best" reading,of course, some critics mean the most valid reading, but the idea ofbestness is widely used to emb.race indiscriminately both the idea ofvalidity and of such aesthetic values as richness, inclusiveness, tension,or complexity-as though validity and aesthetic excellence must some­how be identical.

5

Chapter 1: In Defense of the Author

ing. My demonstration of this point will be found in AppendixI and in the sections on determinacy in Chapter 2. 8 If a theoristwants to save the ideal of validity he has to save the author aswell, and, in the present-day context, his first task will be toshow that the prevailing arguments against the author are

I

questionable and vulnerable.

B. "THE MEANING OF A TEXT CHANGES-

EVEN FOR THE AUTHOR"

A doctrine widely accepted at the present time is that the mean­ing of a text changes. 9 According to the radical historicisticview, textual meaning changes from era to era; according tothe psychologistic view, it changes from reading to reading.Since the putative changes of meaning experienced by theauthor himself must be limited to a rather brief historical span,only the psychologistic view need concern us here. Of course,if any theory of semantic mutability were true, it would legit­imately banish the author's meaning as a normative principlein interpretation, for if textual meaning could change in anyrespect there could be no principle for distinguishing a validinterpretation from a false one. But that is yet another problemthat will be dealt with in a suitable place.lo Here I need notdiscuss the general (and insoluble) normative problems thatwould be raIsed by a meaning which could change, but onlythe conditions that have caused critics to accuse authors of suchfickleness.

8. See pp. 44-48, 225-30.9. See Rene Wellek and Austin Warren, Theory of Literature (New

York, 1948), Chap. 12.10. I have discussed it in Appendix I, pp. 212-16. For the sake of

clarity I should, however, quickly indicate to the reader that verbalmeaning can be the same for different interpreters by virtue of the factthat verbal meaning has the character of a type. A type covers a rangeof actualizations (one example would be a phoneme) and yet in eachactualization remains (like a phoneme) the identical type. This lastpoint is explained in Chap. 2, Sec. D, and in Appendix III, pp. 266-70.

6

B. "The Meaning of a Text Changes"

Everyone who has written knows that his opinion of his ownwork changes and that his responses to his own text vary fromreading to reading. Frequently an author may realize that heno longer agrees with his earlier meaning or expression and willrevise his text. Our problem, of course, has nothing to do withrevision or even with the fact that an author may explain hismeaning differently at different times, since the authors aresometimes inept explainers of their meanings, as Plato ob­served. Even the puzzling case of the author who no longerunderstands his own text at all is irrelevant to our problem,

;:J since his predicament is due to the fact that an author, like any­one else, can forget what he meant. We all know that some­times a person remembers correctly and sometimes not, andthat sometimes a person recognizes his mistakes of memoryand corrects them. None of this has any theoretical interestwhatever.

When critics assert that the author's understanding of histext changes, they refer to the experience that everybody haswhen he rereads his own work. His response to it is different.This is a phenomenon that certainly does have theoretical ill1­portance--though not of the sort sometimes allotted to it.The phenomenon of changing authorial responses is importantbecause it illustrates the difference between textual meaningand what is loosely termed a "response" to the text.

Probably the most extreme examples of this phenomenonare cases of authorial self-repudiation, such as Arnold's publicattack on his masterpiece, Empedocles on Etna, or Schelling'srejection of all the philosophy he had written before] 809. Inthese cases there cannot be the slightest doubt that the author'slater response to his work was quite different from his originalresponse. Instead of seeming beautiful, profound, or brilliant,the work seemed misguided, triVial, and false, and its meaningwas no longer one that the author wished to convey. However,these examples do not show that the meaning of tile work had

. / changed, but precisely the opposite. If the work's meaning had\j changed (instead of the author himself and his attitudes), then

iiii'lL

---------:---- ~_............._"'M' """"'111

1l[

!

C. "It Does Not Matter What an Author Means' ­

that he did not mean anything in particular by his writings.P~es~mably he did mean something by them, and it is a per­mIssIble task to attempt to discover what he meant. Such atask has a determinate object and therefore could be accom- ­plished correctly or incorrectly. However, the task of findinEout what a text says has no determinate object, since the textcan say different things to different readers. One reading is a~

valid or invalid as another. However, the decisive objection toth~ theory o~ semantic autonomy is not that it inconvenientlyfalls to prOVIde an adequate criterion of validity. The decisiveobjection must be sought within the theory itself and in the .faultiness of the arguments used to support it.

One now-famous argument is based on the distinction be­tween a mere intention to do something and the concreteaccomplishment of that intention. The author's desire to com­municate a particular meaning is not necessarily the same ashis suc~ess.in doing so. Since his actual performance is pre- ­sented III hIS text, any special attempt to divine his intention:would falsely equate his private wish with his public accom­plishment. Textual meaning is a public affair. The wide dis­semination of this argument and its acceptance as an axiom ofrecent literary criticism can be traced to the influence of a_vigorous essay, "The Intentional Fallacy," written by W. K. -­Wimsatt and Monroe Beardsley and first published in 1946.1 1 ­

The critic of the arguments in that essay is faced with the :problem of distinguishing between the essay itself and the:popular use that has been made of it, for what is widely takenfor granted as established truth was not argued and could not .have been successfully argued in the essay. Although Wimsatt :and Beardsley carefully distinguished between three types ofintentional evidence, acknowledging that two of them are ,proper and admissible, their careful distinctions and qualifica­tions have now vanished in the popular version which consists

11. Sewanee Review, 54 (1946). Reprinted in William K. Wimsatt, Jr.,The Verbal Icon: Studies in the Meaning of Poetry (Lexington Ky1954). ' .,

10

As I pointed out in section A, this central tenet in the doctrineof semantic autonomy is crucial to the problem of validity. Ifthe tenet were true, then any reading of a text would be "valid,"since any reading would correspond to what the text "says"~for th~t reader. It js useless to introduce normative conceptslike "sensitive," "plausible," "rich," and "interesting," sincewhat the text "says" might not, after all, be any of those things.Validity of interpretation is not the same as inventiveness ofinterpretation. Validity implies the corresp6ndence of an in-"terpretation to a meaning which is represented by the text, andnone of the above criteria for discriminating among interpreta­tions would apply to a text which is dull, simple, insensitive,implausible, or uninteresting. Such a text might not be worthinterpreting, but a crite~ion of validity which cannot cope withsuch a text is not worth crediting.

The proponents of semantic autonomy in England andAmerica can almost always be relied on to point to the exampleof T. S. Eliot, who more than once refused to comment on themeanings of his own texts. Eliot's refusals were based on hisview that the author has no control over the words he hasloosed upon the ~orld and no special privileges as an inter­preter of them. It would have'been quite inconsistent with thisview if Eliot had complained when someone misinterpreted hiswritings, and, so far as I know, Eliot with stoical consistencynever did complain. But Eliot never went so far as to assert'

c. "IT DOES NOT MATTER WHAT AN AUTHOR MEANS~

ONLY WHAT HIS TEXT SAYS"

!hapter 1: In Defense of the Author

ticularly in Chaper 4. For the moment, ~nough has been saidto show that the author's revaluation of his text's significancedoes not change its meaning and, further, that arguments whichrely on such examples are not effective weapons for attackingeither the stability or the normative authority of the author'soriginal meaning.

C. "It Does Not Matter What an Author Means'

that he did not mean anything in particular by his writings.P~es~mablyhe did mean something by them, and it is a per- .mlsslble task to attempt to discover what he meant. Such a'task has a determinate object and therefore could be accom­plished correctly or incorrectly. However, the task of findin!! 'out what a text says has no determinate object, since the tex1 'can say different thi:vgs to different readers. One reading is a~

valid or invalid as another. However, the decisive objection tothe theory of semantic autonomy is not that it inconvenientlyfails to provide an adequate criterion of validity. The decisiveobjection must be sought within the theory itself and in thefaultiness of the arguments used to support it.

One now-famous argument is based on the distinction be­tween a mere intention to do something and the concreteaccomplishment of that intention. The author's desire to com­municate a particular meaning is not necessarily the same ashis success in doing so. Since his actual performance is pre­sented in his text, any special attempt to divine his intention'would falsely equate his private wish with his public accom- .plishment. Textual meaning is a public affair. The wide dis­semination of this argument and its acceptance as an axiom ofrecent literary criticism can be traced to the influence of avigorous essay, "The Intentional Fallacy," written by W. K.Wimsatt and Monroe Beardsley and first published in 1946.11

The critic of the arguments in that essay is faced with theproblem of distinguishing between the essay itself and thepopular use that has been made of it, for what is widely taken;for granted as established truth was not argued and could not;have been successfully argued in the essay. Although Wimsatt ~and Beardsley carefully distinguished between three types of .intentional evidence, acknowledging that two of them areproper and admissible, their careful distinctions and qualifica­tions have now vanished in the popular version which consists

11. Sewanee Review, 54 (1946). Reprinted in William K. Wimsatt, Jr.,The Verbal Icon: Studies in the Meaning of Poetry (Lexington Ky1954). ' .,

C. "IT DOES NOT MATTER WHAT AN AUTHOR MEANS­

ONLY WHAT HIS TEXT SAYS"

Ihapter 1: In Defense of the AuthorJ

iicularly in Chaper 4. For the moment, ~nough has been saidto show that the author's revaluation of his text's significancedoes not change its meaning and, further, that arguments whichrely on such examples are not effective weapons for attackingeither the stability or the normative authority of the author'soriginal meaning.

10

As I pointed out in section A, this central tenet in the doctrineof semantic autonomy is crucial to the problem of validity. Ifthe tenet were true, then any reading of a text would be "valid,"since any reading would correspond to what the text "says"­for that reader. It is useless to introduce normative conceptslike "sensitive," "plausible," "rich," and "interesting," sincewhat the text "says" might not, after all, be any of those things.Validity of interpretation is not the same as inventiveness of

.;C>,-. interpretation. Validity implies the correspondence of an in-', .....~" terpretation to a meaning which is represented by the text, and

Q ;;\~\< none of the above criteria for discriminating among interpreta-t}J- tions would apply to a text which is dull, simple, insensitive,

implausible, or uninteresting. Such a text might not be worthinterpreting, but a crite~ion of validity which cannot cope withsuch a text is not worth crediting.

The proponents of semantic autonomy in England andAmerica can almost always be relied on to point to the exampleof T. S. Eliot, who more than once refused to comment on themeanings of his own texts. Eliot's refusals were based on hisview that the author has no control over the words he hasloosed upon the »,orld and no special privileges as an inter­preter of them. It would have been quite inconsistent with thisview if Eliot had complained when someone misinterpreted hiswritings, and, so far as I know, Eliot with stoical consistencynever did complain. But Eliot never went so far as to assert

j.

I:,;~

I,

------------------------ .~".

13

C. "It Does Not Matter What an Author Means"

convey desolation, and if to every competent reader his poemconveyed only a sense that twilight is approaching, then sucLpublic unanimity would make a very strong case (in this par··ticular instance) for the practical irrelevance of the author'sintention. But when has such unanimity occurred? If it existedgenerally, there would not be any problems of interpretation.

The myth of the public consensus has been decisive in gain·· •ing wide acceptance for the doctrine that the author's intention \is irrelevant to what the text says. That myth permits the con- i,

fident belief that the "saying" of the text is a public fact firmlYt'governed by public norms. But if this public meaning exists,why is it that we, who are the public, disagree? Is there onegroup of us that constitutes the true public, while the rest areheretics and outsiders? By what standard is it judged that acorrect insight into public norms is lacking in all those readerswho are (except for the text at hand) competent readers oftexts? The idea of a public meaning sponsored not by theauthor's intention but by a public consensus is based upon afundamental error of observation and logic. It is an empiricalfact that the consensus docs not exist, and it is a logical errorto erect a stable normative concept (i.e. the public meaning)out of an unstable descriptive one. The public meaning of atext is nothing more or less than those meanings which thepublic happens to construe from the text. Any meaning whichtwo or more members of the public construe is ipso' facto with­in the public norms that govern language and its interpretation.Vox populi: vox populi.

If a text means what it says, then it means nothing in par­e- I ticular. Its saying has no determinate existence but must be,.. the saying of the author or a reader. The text does not exist

even as a sequence of words until it is construed; until then, itis merely a sequence of signs. For sometimes words can havehomonyms (just as, by analogy, entire texts can), and some­times the same word can b~ quite a different word. For ex­ample, when we read in Wordsworth's Intimations Ode thephrase "most worthy to be blessed," are we to understand

Chapter 1: In Defense of the Author

in the false and facile dogma that what an author intended isirrelevant to the meaning of his text.

The best way to indicate what is fallacious in this popularversion is to discuss first the dimension in which it is perfectlyvalid---evaluation. It would be absurd to evaluate the stylisticfelicity of a text without distinguishing between the author's in­tention to convey a meaning and, on the other hand, his effec­tiveness in conveying it. It would be similarly absurd to judgethe profundity of a treatise on morality without distinguishingbetween the author's intention to be profound and his successin beings<:;-: EvaluatIon is constantly distinguishing between in­tention and accomplishment. Take this example: A poet in­tends in a four-line poem to convey a sense of desolation, butwhat he manages to convey to some readers is a sense that thesea is wet, to others that twilight is approaching. Obviously hisintention to convey desolation is not identical with his stylisticeffectiveness in doing so,' and the anti-intentionalists quitejustly point this out. But the intentional fallacy is properly

Vi, applicable only to artistic success and to other normative cri­teria like profundity, consistency, and so on. The allti-il1ten­tionalist quite properly defends the right and duty of the criticto judge freely on his own criteria and to expose discrepanciesbetween wish and deed. However, the intentional fallacy hasno proper appllcation whatever to verbal meaning. In theabove example the only universally valid meaning of the poemis the sense of desolation. If the critic has not understood thatpoint, he will not even reach an accurate judgment-namely,that the meaning was ineptly expressed and perhaps was notworth expressing in the first place.l 2

Beneath the so-called intentional fallacy and, more generally,the doctrine of semantic autonomy lies an assumption which iftrue would at least render plausible the view that the meaningof a text is independent of its author's intention. I refer to theconcept of a public consensus. If a poet intended his poem to

12. For a definition of verbal meaning see Chap. 2, Sec. A.

'.'...

Chapter 1: In Defense of the Author

"most" as a superlative or merely an intensifier like "very"?Even on this primitive level, signs can be variously construed,and until they are construed the text "says" nothing at all.

)

D. "THE AUTHOR'S MEANING IS INACCESSIBLE"

Since we are all different from the author, we cannot reproducehis intended meaning in ourselves, and even if by some ac­cident we could, we still would not be certain that we had doneso. Why concern ourselves, th~refore, with an inherently im­possible task when we can better employ our energies in usefuloccupations such as making the text relevant to our presentconcerns or judging its conformity to high standards of excel­lence? The goal of reproducing an inaccessible and private pastis to be dismissed as a futile enterprise. Of course, it is essentialto understand some of the public facts of language and historyin order not to miss allusions or mistake the contemporarysenses of words, but these preliminary tasks remain squarelyin the public domain and do not concern a private worldbeyond the reach of written language.

Before touching on the key issue in this argument-namely,that the author's intended meaning cannot be known-I wouldlike to make an observation about the subsidiary argumentrespecting the public and private dimensions of textual mean­ing. According to this argument, it would be a mistake to con­fuse a public fact-namely, language-with a private fact­namely, the author's mind. But I have never encountered aninterpretation that inferred truly private meanings from a text.An interpreter might, of course, infer meanings which accord­ing to our judgment could not possibly under any circum­stances be implied by the author's words, but in that case, wewould reject the interpretation not because it is private but be­cause it is probably wrong. That meaning, we say, cannot beimplied by those words. If our skepticism were shared by allreaders of the interpretation, then it would be reasonable to say

14

I'"~

D. "The Author's Meaning Is Inaccessible"

that the interpretation is private. I-Iowever, it is a rare inter­pretation that does not have at least a few adherents, and if ithas any at all, then the meaning is not private; it is at worstimprobable.

Whenever an interpretation manages to convince anotherperson, that in itself proves beyond doubt that the author'swords can publicly imply such a meaning. Since the interpretedmeaning was conveyed to another person, indeed to at leasttwo other persons, the only significant interpretive question is,"Did the author really intend that public meaning by hiswords?" To object that such a meaning is highly personal andought not to have been intended is a legitimate aesthetic ormoral judgment, but is irrelevant to the question of meaning.

>:tfThat meaning-if the author did mean it-has proved itself to, I be public, and if the interpreter manages to do his job con­

\ vincingly, the meaning can become available to a very largeI public. It is simply a self-contradiction for a member of the[ public to say, "Yes, I see that the author did mean that, but it

is a private not a public meaning."The impulse that underlies this self-contradictory sor1 of

argument is a sound insight that deserves to be couched interms more suitable than "public" and "private." The issue isfirst of all a moral and aesthetic one. It is proper 10 demand ofauthors that they show consideration for their readers, that theyuse their linguistic inheritance with some regard for the gen­erality of men and not just for a chosen few. Yet many newusages are bound to elude the generality of men until readersbecome habituated to them. The risk of resorting to semi­private implications-available at first only to a few-is veryoften worth taking, particularly if the new usage does finallybecome widely understood. The language expands by virtueof such risky innovations. However, the soundest objection toso-called private meanings does not relate to moral and aesthet­ic judgment but to the practice of interpretation. Those inter­preters who look for personal implications in such formalizedutterances as poems very often disregard genre conventions

15

13. See Chap. 2, Sec. B, and Chap. 4, Sec. A and B.

16

D. "The Author's Meaning Is Inaccessible"

because it is self-evidently true. I can never know anotherperson's intended meaning with certainty because" I cannot getinside his head to compare the meaning he intends with themeaning I understand, and only by such direct comparisoncould I be certain that his meaning and my own are identical.But this obvious fact should not be allowed to sanction theoverly hasty conclusion that the author's intended meaning isinaccessible and is therefore a useless object of interpretation.!U~£!.l~gicalmistake to confuse the impossibility of certainty vinYE~.erstanaiii~~ith the impossibility ofuIlde_rst,!nding.U"isasimilar;Thougll-more"s"ubtle,-ffilsOfa-keOta-identify knowledge withcertainty. A good many disciplines do not pretend to certainty,and the more sophisticated the methodology of the discipline,the less likely that its goal will be defined as certainty of knowl­edge. Since genuine certainty in interpretation is impossible,the aim of the discipline mlist be to reach? consensus, on the v"(basis of what is known, that correct understa~gh~sprobablybeen achieved. The issue is not whether certainty is accessibleto the interpreter but whether the author's intended meaningis accessible to him. Is correct understanding possible? That isthe question raised by the thesis under examination.

Most of us would answer that the author's meaning is onlypartially accessible to an interpreter. We cannot know all themeanings the author entertained when he wrote down his text,as we infer from two familiar kinds of evidence. Whenever Ispeak I am usually attending to ("have in mind") meanings thatare outside my subject of discourse. Furthermore, I am alwaysaware that the meanings I can convey through discourse aremore limited than the meanings I can entertain. I cannot, forexample, adequately convey through words many of my visualperceptions-though these perceptions are meanings, which isto say, objects of consciousness. It is altogether likely that notext can ever convey all the meanings an author had in mind ashe wrote.

But tbis obvious fact is not decisive. Why should anyonewith common sense wish to equate an author's textual meaning

17

, ':,i

Chapter 1: In Defense of the Author

'with all the meanings he happened to entertain when he wrote?Some of these he had no intention of conveying by his words.

\ Any author kno",,:s that writ~en verbal utte~ances c~n conveyonly verbal meamngs-that IS to say, meanmgs WhIch can be

\ conveyed to others by th~ word~ he uses.J~~J~~te;p~~t<l!.~~of\ texts is<::o~cernedexclusI,\,~lY__\~llth ~!Iar.~!>J~JTl~,<!!l.!.Ilg~LClIl~Lnotre~erythiI1KI?.llijlJiJj.IcJrig of w hen~ I ""rite .c;.<l~ be _~~~~.9_~J!hlothers'bY meansgflpy wor21s~'coiiverseiy,many of my shar-

7 iJ:ble-~ea~ings' are meiJ:ningswhich I am not directly thInkingV of at all. They are so~calledunconscious meanings.l 4 It betrays

a totally inadequate conception of verbal meaning to equate itwith what the author "has in mind." The only question that canrelevantly be at issue is whether the verbal meaning which anauthor intends is accessible to the interpreter of his text.

Most authors believe in the accessibility of their verbal mean­ing, for otherwise most of them would not write. However, noone could unanswerably defend this universal faith. Neither theauthor nor the interpreter can ever be certain that communica­tion has occurred or that it can occur. But again, certainty isnot the point at issue. It is far more likely that an author and:an interpreter can entertain identical meanings than that they I

cannot. The faith that speakers have in the possibility of com~munication has been built up in the very process of learning alanguage, particularly in those instances when the actions of theinterpreter have confirmed to the author that he has beenunderstood. These primitive confirmations are the foundationfor our faith in far less primitive modes of communication. Theinaccessibility of verbal meaning is a doctrine that experiencesuggests to be false, though neither experience nor argumentcan prove its falsity. But since the skeptical doctrine of in~

accessibility is highly improbable, it should be rejected as aworking assumption of interpretation.

Of course, it is quite reasonable to take a skeptical positionthat is less sweeping than the thesis under examination: certain

14. See Chap. 2, Sees. D and E.

18

~._,--,--------

~i

E. "The Author Often Does Not Know What He Means"

texts might, because of their character or age, represent au­thorial meanings which are now inaccessible. No one would,I think, deny this reasonable form of skepticism. However,similar versions of such skepticism are far less acceptable, par­ticularly in those theories which deny the accessibility of theauthor's meaning whenever the text descends from an earliercultural era or whenever the text happens to be literary. Theseviews are endemic respectively to radical historicism and to thetheory that literary texts are ontologically distinct from non­literary ones. Both of these theories are challenged in sub­sequent chapters. However, even if these theories were accept­able, they could not uphold the thesis that an author's verbalmeaning is inaccessible, for that is an empirical generalizationwhich neither theory nor experience can decisively confirm ordeny. Nevertheless, with a high degree of probability, that gen­eralization is false, and it is impossible and quite unnecessaryto go beyond this conclusion.

E. "THE AUTHOR OFTEN DOES NOT KNOW

WHAT HE MEANS"

Ever since Plato's Socrates talked to the poets and asked themwith quite unsatisfactory results to explain "some of the mostelaborate passages in their own writings," it has been a com­monplace that an author often does not really know what hemeans.l 5 Kant insisted that not even Plato knew what hemeant, and that he, Kant, could understand some of Plato'swritings better than Plato did himself. 16 Such examples of

15. Plato, Apology, 22b-c.16. Immanuel Kant, Critique of Pure Reason, trans. N. K. Smith

(London, 1933), A 314, B 370, p. 310: "Ishall not engage here in anyliterary enquiry into the meaning which this illustrious author attachedto the expression. I need only remark that it is' by no means unusual,upon comparing the thoughts which an author has expressed in regardto his subject, whether in ordinary conversation or in writing, to findthat we understand him better than he has understood himself."

19

"I'

,,: ;

: f

ci

I\, \

Chapter 1: In Defense of the Author

authorial ignorance are, no doubt, among the most damagingweapons in the attack on the author. If it can be shown (as itapparently can) that in some cases the author does not reallyknow what he means, then it scems to follow that the author'smeaning cannot constitute a general principle or norm fordetermining the meaning of a text, and it is precisely such ageneral normative principle that is required in defining theconcept of validity.

Not all cases of authorial ignorance are of the same type.Plato, for instance, no doubt knew very well what he meant byhis theorY of Ideas, but it may have been, as Kant believed,that the theory of Ideas had different and more general impli­cations than those Plato enunciated in his dialogues. ThoughKant called this a case of understanding the author better thanthe author understood himself, his phrasing was inexact, ror itwas not Plato's meaning that Kant understood better thanPlato, but rather the subject matter that Plato was attemptingto analyze. The notion that Kant's understanding of the Ideaswas superior to Plato's implies that then~_is a subject matterto which Plato's meaning was inadcquate.,)If we do not makethis distinction between subject mattcr and meaning, we haveno basis for judging that Kant's understanding is better thanPlato's'! 7 Kant's statement would have been more precise ifhe had said that he understood the Ideas better than Plato,not that he underst~odPlato's meaning better than Plato. If wedo not make, and preserve the distinction between a man'smeaning and his subject matter, we cannot distinguish betweentrue and false, better and worse meanings.

This example illustrates one of the two main types of au­thorial ignorance.iIt has greatest importance in those genresof writing that aspire to tell the truth about a particular subjectmatter. The other principal type of authorial ignorance pertainsnot to the subject matter but to the author's m~i~gjtself,and '(

17. The distinction between meaning and subject matter is discussedin Chap. 2, Sec. F, and is one foundation for my objections to Gadamer'sidentification of meaning with Sache. See Appendix II, pp. 247-49.

20

---------_..' ....... ~."

E. "The Author Often Does Not Know What He Means"

can be illustrated whenever casual conversation is subjecte0to stylistic analysis: '

"Did you know that those last two sentences of yourshad parallel constructions which emphasized their sim.. 'ilarity of meaning?"

"No! Bow clever of mel I suppose I really did want toemphasize the similarity, though I wasn't aware of that",and I had no idea I was using rhetorical devices to do it." \

What this example illustrates is that there arc usually com- \\.ponents of an author's intended meaning that he is not con- Jscious of. It is precisely here, where an interpreter makes theseintended but unconscious meanings explicit, that he can right­fully claim to understand the author better than the authorhimself. But here again a clarification is required. The inter­preter's right to such a claim exists only when he carefullyavoids confusing meaning with subjcct matter, as in the ex­ample of Plato and Kant. The interpreter may believe that heis drawing out implications that are "necessary" accompani­ments to the author's meaning, but such necessary accom­paniIncnts arc rarely unavoidable components of someone'smeaning. They become necessary associations only within agiven subject matter.18 For example, although the concept"two" necessarily implies a whole array of concepts includingthose of succession, integer, set, and so on, these may not beimplied in a given usage of the word, since that usage could beinadequate or misconceived with respect to the subject mattcrin which "two" falls. Only within that subject mattcr does theresubsist necessity of implication. Thus, by claiming to perceiveimplications of which the author was not conscious, we maysometimes distort and falsify the meaning of which he wasconscious, which is not "better understanding" but simplymisunderstanding of the author's meaning.

18. This distinction was not observed in the interesting essay byO. Bollknow, "Was heisst es einen Verfasser zu verstehen besser als ersich seIber verstanden hat?" in Das Verstehen, Drei Aujsiitze zurTheorie des Geisteswissenschajten (Mainz, 1949).

21

Ii.,...,.~;.

Gl2&W"'MY"r

,0'J

Chapter 1: In Defense of the Author

But let us assume that such misunderstanding has beenavoided and that the interpreter really has made explicit certainaspects of an author's undoubted meaning of which the authorwas unconscious-as in stylistic analysis of casual conversa­tion. The further question then arises: How can an author meansomething he did not mean? The answer to that question issimple. It is not possible to mean what one does not mean,though it is very possible to mean what one is not conscious

~/v J of meaning. That is the entire issue in the argument based onauthorial ignorance. That a man may not be conscious of allthat he means is no more remarkable than that he may not beconscious of all that he does. There is a differen~~_ be!ween

/-' meaning and consciousness of me'aI;ing, and since meaning isan affair of consciousness,one cansaY"more'preclSe1ythatl.nere-is a difference between.c~nsciQusness.~~ii5Ls~~lf=E§ris~2.1.ff'ness::i:~deed, when an author's meaning is complicated, heca~~ot possibly at a given moment be paying attention to allits complexities. But the distinction between attended and un­attended meanings is not the same as the distinction betweenwhat all author means and what he does not mean. H ! Noexample of the author's ignorance with respect to his meaningcould legitimately show that his intended meaning and themeaning of histext are two different things.

Other varieties 'of authorial ignorance are therefore of littletheoretical interest. When Plato observed that poets could notexplain what they meant, he intimated that poets were in­effectual, weak-minded, and vague-particularly with respectto their "most elaborate passages." But he would not have con­tended that a vague, uncertain, cloudy, and pretentious mean­ing is not a meaning, or that it is not the poet's meaning. 2o

Even when a poet declares that his poem means whatever it istaken to mean (as in the case of some modern writers who

19. For a discussion of so-called conscious and unconscious mean­ings see Chap. 2, Sec. D and E.

20. Or at least that of the muse who temporarily possesses him­the muse being, in those unseemly cases, the real author.

22

E. "The Author Often Does Not Know What He Means"

believe in the current theory ~f public meaning and autho~ial (irrelevance), then,-?o~c:l,~?~-'lll~poe~~aXllot1l!~a~ al1~,thl~g \in particular. Yet eVen III such a hmltmg case It IS stIll theauthoi'who'''determines'' the meaning,

One final illustration of authorial ignorance, a favoriteamong literary critics, is based on an examination of anauthor's early drafts, which often indicate that what the authorapparently intended when he began writing is frequently quitedifferent from what his final work means. Such examples showhow considerations of style, genre, and local texture may playa larger part in his final meaning than that played by hisoriginal intention, but these interesting observations havehardly any theoretical significance. If a poet in his first draftmeans something different than he means in his last, it doesnot imply that somebody other than the poet is doing themeaning. If the poet capitalizes on a local effect which he hadnot originally intended, so much the better if it makes a betterpoem. All this surely does not imply that an author does notmean what he means, or that his text docs not mean what heintends to convey.

If there is a single moral to the analyses of this chapter, it isthat meaning is an affair of consciousness and not of physicalsigns or things. Consciousness is, in turn, an affair of persons,and in textual interpretation the persons involved are an authorand a reader. The meanings that are actualized by the readerare either shared with the author or belong to the reader alone.While this statement of the issue may affront our deeply in­grained sense that language carries its own autonomous mean­ings, it in no way calls into question the power of language. O~the contrary, it takes for granted that all meaning commum­cated by texts is to some extent language-bound, that no textualmeaning can transcend the meaning possibilities and the con­trol of the language in which it is expressed. What has beendenied here is that linguistic signs can somehow speak theirown meaning-a mystical idea that has never been persuasivelydefended.

23

2.MEANING AND IMPLICATION

"The question is," said Alice, "whether youcan make words mean so many different things."

"The question is," said Humpty Dumpty, "whichis to be master-that's all."

Lewis Carroll

lSince it is very easy for a reader of any text to construe mean­ings that are different from the author's, there is nothing in thenature of the text itself which requires the reader to set up theauthor's meaning as his nonnative ideal. Any normative con­cept in interprctation implies a choice that is required not bythe nature of written texts but rather by the goal that the inter­preter sets himself] It is a weakness in many descriptions ofthe interpretive process that this act of choice is disregardedand the process' described as though the object of inter­pretation were somehow determined by the ontological statusof texts themselves. The argument, for example, that-changingcultural conditions change the meaning of a text assumes thatthe object of interpretation necessarily changes under changedconditions. Similarly, the defense of re-cognitive interpretation \often assumes that something in the nature of a text requiresthe meaning to be the stable and determinate meaning of anauthor. 1 But the object of interpretation is precisely that which

\i 1. I borrow the term "re-cognitive interpretation" from EmilioBetti. A re-cognition implies, of course, the cognition of what theauthor had cognized (i.e. meant)-Boeckh's "Erkennen des Erkannten."Although the term here embraces wider domains than are usually in­cluded under "cognition" (i.e. unconscious and emotive domains), the

24

Chapter 2: Meaning and Implicatir

cannot be defined by the ontological status of a text, since fl':.distinguishing characteristic of a text is that from it not jl1~'

one but many disparate complexes of meaning can be costrued. ,Only by ignoring this fact can a theorist attempt to err- ~.

a normative principle out of a neutral and variable state oJ:affairs-a fallacy that seems endemic to discussions of he,-­meneutics. Bluntly, no necessity require~ the object of inte Co.

pretation to be determinate or indeterminate, changing or un..changing. On the contrary, the object of interpretation is 110

automatic given, but a task that the interpreter sets himse1LHe decides what he wants to actualize and what purpose hisactualization should achieve.

Thus, while it is a fallacy to claim that a particular norm forinterpretation is necessarily grounded in the nature of this orthat kind of text, rather than in the interpreter's own will, it isquite another matter to claim that there can be only one sort ofnorm when interpretation is conceived of as a corporate enter­prise. For it may very well be that there exists only one normthat can be universally compelling and generally sharable. Inthe previous chapter I argued that no presently known norm,]­tive concept other than the author's meaning has this univcr-­sally compelling character. On purely practical grounds, there..fore, it is preferable to agree that the meaning of a text is theauthor's meaning.

Usually it is true that the defense of the old ideal of re­cognitive interpretation is carried out on a different front. It ispointed out that the main reason for studying texts, particularlyold ones, is to expand the mind by introducing it to the im­mense possibilities in human actions and thoughts-to see andfeel what other men have seen and felt, to know what theyhave known. Furthermore, none of these expansive benefitscomes to the man who simply discovers his own meanings in

sympathetic reader will make the appropriate adjustment. See EmilioBetti, Teoria generale della interpretazione (2 vols. Milan, Giuffre,1955),1,343--432.

25

Chapter 2: Meaning and Implication

someone else's text and who, instead of encountering anotherperson, merely encounters himself. When a reader does that,he finds only his own preconceptions, and these he did not needto go out and seek. Finally, the defender of re-cognitive inter­pretation adds that the knowledge of what has been thoughtand felt is also, after all, a form of knowledge and, as such,worth gaining for its own sake.

There is nothing despicable iri this argument, nor can anyconsiderable objection be raised against it, except that theknowledge sought may be, for various reasons, impossible toachieve. Some of these skeptical objections I have alreadyanswered in principle, and in this chapter I shall deal with thetwo root forms of all such skepticism-psychologism and radi-

( cal historicism. However, I shall not repeat at length the moral\ arguments in favor of viewing interpretation as are-cognition

vI of the author's meaning.It is, of course, quite true th'at the choice of a norm for

interpretation is a free social and ethical act. Any reader canadopt or reject any norm, and he is justified in thinking thatthere is no absolute necessity for his choosing one or another.Furthermore, he mayor may not accept the idea that all usesof language carry moral imperatives which .derive from thedouble-sided, interpersonal character of linguistic acts. Allthis he may reject as unconvincing, and nothing in the mutesigns before him will compel him to change his mind or bringhim ill fortune if he does not. Partly for this reason, I pavechosen a different sort of defense--one that appeals not tothe ethICS of language but to the logical consequences that

-Y follow from the act of public interpretation. As soon as anyoneclaims validity for his interpretation (and few would listen to acritic who did not), he is immediately caught in a web of logicalnecessity. If his claim to validity is to hold, he must be willingto measure his interpretation against a genuinely discriminatingnorm, and the only compelling normative principle that has

j ever been brought forward is the old-fashioned ideal of rightlyunderstanding what the author meant. Consequently, my case

26

. , ..." - ~ '"' "- ... ~.

A. Defining V erbal Meaning

rests not on the powerful moral arguments for re-cognitiveinterpretation, but on the fact that it is the only kind of inter- ?pretation with a determinate object, and thus the only kindthat can lay claim to validity in any straightforward and prac­ticable sense of that term.

While the problem of validity is consistently circumventedby those who attack the ideal of re-cognitive interpretation,the substantial elements in their attack cannot be ignored. Eventhough only one compelling normative principle exists, it isstill necessary to show that it is a viable principle. Thus, I shallhave to show that the author's verbal meaning is determinate,that it is reproducible, and finally that it provides a means forcoping with the knottiest problem of interpretation, the problemof implication.

Such an account of the authorial meaning that is sought byre-cognitive interpretation should serve as a foundation for allother interpretive goals as well. For even when the originalauthor is rejected or disregarded, any construction of a textstill constitutes a meaning that must have an author-thoughhe be merely the critic himself. All forms of written interpreta­tion and all interpretive goals that transcend private experiencerequire that some author's meaning be both determinate andreproducible. In discussing the nature of verbal meaning, I shallpay particular attention to these two universal requirements.

A. DEFINING VERBAL MEANING

Although verbal meaning requires the determining will of anauthor or interpreter, it is nevertheless true that the norms oflanguage exert a powerful influence and impose an unavoidablelimitation on the wills of both the author and interpreter. Aliceis right to say that Humpty Dumpty cannot successfully makewords mean just anything he wants them to. Therefore, anydiscussion' of verbal meaning should define in principle theway in which linguistic norms exert this codetermining in­fluence.

27

d· ,.

.. -',

Chapter 2: Meaning and Implication

Is language always constitutive of verbal meaning or is itsometimes merely a controlling factor that sets limits to pos­sible verbal meanings? This problem has been much discussed,and like many othcrs in the purview of hermeneutics, it prob­ably cannot be solved with certainty, since no satisfaCtory wayof testing either hypothesis has been devised. Ncvertheless, itis very probable that neither hypothesis is true for all instancesand sorts of verbal meaning. Sometimes a use of language isuniquely constitutive of meaning; sometimes, apparently, aparticular choice of words merely imposes limitations and isnot uniquely required for the meaning that is actually willed.This is suggested by the example of translation. Some utter­ances, particularly of a technical sort, can be perfectly trans­lated, while others, particularly lyric poems, are never per­fectly carried over into another language. It seems to followthat the language-bound quality of utterances, that is, thedegree to which language is constitutive of meaning, can varyfrom null to somewhere in the vicinity of 100 per cent.

Certainly the claims of the meta-linguists and the proponentsof Muttersprache seem to be far too absolute. They have given~onvincing and impressive examples of the way language canconstitute thought and meaning and have reminded us thatHumboldt's conceptiQI1 of language as energeia was an epocli­making insight. But these observations do not compel theunprovable and improbable conclusion that a unique use oflanguage is always constitutive of a unique meaning;2 Theargument that a Muttersprache imposes an inescapable Welt­anschauung on its speakers seems to overlook the remarkablevariety in the assumptions and attitudes of speakers who havethe same Muttersprache. On this point the sagest commentI have encountered is by Manfred Sandmann:

It would be wrong to infer from the absence of an ade­quate linguistic sign an ignorance of the correspondingthing-meant (German has no word for bully, English

2. See Chap. 3, Sec. E.

28

A. Defining Verbal Meaning

no word for Schadenfreude); it would be equally wrongto conclude that an English-speaking person could not seethat dew, rain, ice, water, mist, etc., were only differentstates of the same thing simply because there is no wordfor that thing in the English language. 3

An amusing illustration of Sandmann's very sensible posi­tion is found in a contest set by Paul Jennings in The Observerof 20 December 1964. The contestant was to invent a wordfor each of ten definitions, such as "to make a sound likeescaping bath water," "to pursue an excessive standard of liv­ing or 'keep up with the Joneses,' " and "having the appearanceof affluence but living on credit." Obviously meanings such asthese frequently exist before individual words become gen­erally available to express them, and it is, of course, true thatwhen such words do become available they may alter (i.e.partly constitute) the meanings they were devised to express.For example, a single word, by virtue of its compactness, mayhave quite a different effect than the definition it was designedto support. This concentrating and hypostatizing effect may bevery important in some utterances but may carry no meaningvalue at all in others. No absolute, a priori pronouncement iswarranted, sinee the effect mayor may not be operative in aparticular utterance. This rejection of absolute, a priori gen­eralizations with regard to linguistic effects that are variableand local is one of the main points of this book.4

On the other hand, it obviously is warranted to say thatlinguistic norms at the very least always impose limitations onverbal meaning. In the first place, there exist limitations whichare intrinsic to all linguistic media. For example, it is impos­sible to express in language meanings that are constituted by amedium that is not linguistic, such as music or painting. 5 How-

3. Manfred Sandmann, Subject and. Predicate: A Contribution tothe Theory of Syntax (Edinbmgh, 1954), p. 73.

4. See partiCUlarly Chap. 3, Sees. B-E.5. Of course language often refers to meanings constituted by other

media, though it cannot aC;;curately translate those meanings.

29

Chapter 2: M eatzing and Implication

ever, this general limitation on the possibilities of language isnot of great significance in textual interpretation, since itmerely states the tautology that what is constituted by anothermedium is constituted by it and therefore cannot be separatedfrom it. Of far greater importance in hermeneutical the,ory arethose meanings excluded from language not by their naturebut by the linguistic norms that actually obtain. The operationof these limitations may be called for convenience the Humpty­Dumpty effect. Although Saussure is convincing when heargues that the potentialities of a language are finite at anymoment in time, these linguistic limitations can never belegislated in advance. The most important version of theHumpty-Dumpty effect is the one that Alice pointed out: whensomebody does in fact use a particular word sequence, hisverbal meaning cannot be anything he might wish it to be.This very general restriction is the single important one forthe interpreter, who always confronts a particular sequence oflinguistic signs.

Yet even in confronting a particular word sequence theinterpreter rnust recognize that "the norms Qf language" are·not a uniform set of restrictions, requirements, and patterns ofexpectation, but an immense number of different ground rulesthat vary greatly with respect to different utterances. This pointhas been made with great clarity and concreteness by Wittgen­stein. 6 The generalization that "mind reading" is not the dutyof the interpreter, who is obliged to understand only thosemeanings which "the public norms of language" permit, mayplausibly apply to many kinds of formal discourse, but it wouldnot apply to a parent's interpretation of a child's ellipticalstatements or to the frequent ellipses of ordinary conversation.Since these utterances do carry verbal meaning under the par­ticular norms that obtain for such speech acts, the principlethat verbal meaning is limited by the norms of language doesnot constitute any easy a priori narrowing of the interpreter's

6. In PhilosophicaL Investigations, trans. G. E. M. Anscombe (NewYork, 1953), p. 26, and passim.

30

B. Reproducibility: Psychologistic Objections

task. The norms of language are neither uniform nor stable butvary with the particular sort of utterance that is to be inter­preted.

A single principle underlies what we loosely call "the normsof language." It is the principle of sharability. Because shar­ability is the decisive element in all linguistic norms, it is im­portant to conceive of them, despite their complexity andvariability, on this fundamental level. We thereby place em­phasis not on the structural characteristics of the linguisticmedium, but on the function of speech, which is our centralconcern. Theory of interpretation need not and ought notdescribe linguistic norms merely in terms of syntax, grammar,meaning kernels, meaning fields, habits, engrams, prohibitions,and so on, all of which are extremely variable and probablyincapable of adequate description. It is more important toemphasize the huge and unencompassable areas of meaning­including emotional and attitudinal meanings-that languageactually does represent. Considering this immensity taken as awhole, the restrictions imposed by all the different varieties oflingoistic ground rules do not require special emphasis. It isby no means a denial of those restrictions to say that thecapacity of language to represent all conceivable meanings isultimately limited only by the overarching principle of shar­ability.

On these grounds I offer the following as a provisional,concise definition of verbal meaning, to be expanded andexplored later in this chapter: Verbal meaning is whateversomeone has willed to convey by a partiCUlar sequence oflinguistic signs and which can be conveyed (shared) by meansof those linguistic signs.

B. REPRODUCIBILITY: PSYCHOLOGISTIC OBJECTIONS

The reproducibility (and thus the sharability) of verbal mean­ing depends on there being something to reproduce. For themoment I will assume that any verbal meaning as defined above

31

~.'

THE PSYCHOLOGUS

Psychologus

Characters:

Socrates

s. Here he comes now. Hello, Psychologus. We were justtalking about you.

P. Well, if it isn't Socrates! How have you been after allthis time? You're looking well. Haven't changed a bitsince we last met some months ago-though you reallyhave changed, of course, since everybody does. Takeme, for example. I'm not the same as I was when welast met. My feelings and experiences are different, andfrankly I'm older.

S. You are a philosopher, Psychologus. I hadn't beenthinking of such high things at all. In fact, I was saying

B. Re.producibility: Psychologistic Objection':

hermeneutics, no decisive answer can possibly be given. Thepsychologistic notion that one man's meaning is always dif­ferent from another's is not an empirical theory that can ever befalsified by an empirical test, since no one can ever be certainprecisely what meaning another man entertains.

Nevertheless, the psychologistic conception can be shownto be inadequate as a theory of meaning, because it is notcapable of explaining how quite different mental processes canproduce an identical meaning, and that is an experience whichoccurs consistently in the mental processes of one and thesame person at different moments of time. Thus, while it couldnever be shown that two different persons entertained identicalmeanings, it can be shown that the psychologistic theory ofmeaning is wrong. A far better argument is required to upholdthe view that an interpreter's meaning is always necessarilydifferent from an author's. The inadequacy of that argumentwas exposed centuries ago in a Platonic dialogue which I re­produce in its entirety, being convinced that Socrates' ironypoints up the issue far more effectively than my sober exposi­tion ever could.

Chapter 2: Meaning and Implication

is a determinate entity with a boundary that discriminates whatit is from what it is not. I shall discuss the nature of thatboundary in the last part of this chapter, but first I shall con­sider the objections raised by those who deny the possibilityof re-cognitive interpretation on the grounds that verbal mean­ing is never perfectly reproducible. The most widespread ob­jection is that an interpreter by necessity must understand ameaning that is different from the author's meaning becausethe interpreter is different from the author. This objection holdseven when the interpreter is the author himself, since no manis precisely the same at different ~imes.

The argument that an interpreter's understanding is neces­sarily different because he is different assumes a psychologisticconception of meaning which mistakenly identifies meaningwith mental processes rather than with an object of thoseprocesses. Since (the argument runs) an interpreter's experi­ences, feelings, attitudes, habitual responses are all differentfrom the author's, so therefore must be the meanings that heconstrues from the words before him. Since one man's con­ception of and response· to a rainbow is always different insubtle ways from another man's, so therefore must be hisunderstanding of the word "rainbow." I say that this viewequates meaning with mental processes because the undoubtedfact that one man's mental life is not the same as another's isthought to be a sufficient ground for concluding that he under­stands different meanings. If something different is going onin his head, then what he understands has to be different. Thus,under this conception, meaning is in effect identified with aparticular complex of mental acts.

That an interpreter can and often does fail to understandexactly what an author means by a word like "rainbow" is notin dispute. Nor would a sensible man deny that such misunder­standings are frequently caused by the fact that one man'sconception of and response to a rainbow are different from an­other's. The crucial question is, Must such misunderstandingsnecessarily occur? But to this question, like so many others in

32

(::,,f

Chapter 2: Meaning and Implication

that you were able to perceive the subtlest differencesin meaning every time you encountered the samewords. We had been talking about the word "rainbow."

P. Absolutely right. I'm sure you understand that myonly interest in making that point is to g'et the matterstraight and help people to get rid of their naIve illu­sions. Actually there is a little poem about a rainbow~you know: "My heart leaps up," and so on~and Ican tell you quite frankly, Socrates, that for me it is adifferent poem every time I read it.

S. It means something different to you every time?P. Precisely. As I was saying, I'm different myself every

time, and I have different associations and responses.Entre nous, I used to like it but now more often thannot it leaves me cold.

S. It is not now as it hath been of yore?P. Ah! I see you read poetry too. That's quite a change

for you, Socrates.S. Yes, J think that's your point about people changing.

But 1 am troubled by something~thoughI'm not sureprecisely what it is. It has to do with your saying thatit's a different poem every time you read it, while Ithought I also understood you to say that it was alwaysthe same poem that you read.

P. Socrates, it's hard to decide sometimes whether youare being sly or just simpleminded. When -we say it'sthe same poem, that is just a loose manner of speaking.The poem isn't the'same at all. We call it the same forconvenience because the words are the same everytime even though the meaning isn't.

S. You mean the physical signs stay the same, thoughwhat they mean changes?

P. Precisely.S. No, I don't think that's quite the way to put it, because

I'm wondering whether we really ought to call the signs

the same.

34

B. Reproducibility: Psychologistic Objections

P. Why not?S. Well, sometimes I might read the poem in another

book or even in a manuscript, so the physical signswould be different even though I called the poem thesame. I don't think it can be the physical signs thatare the same.

P. You do like to stretch things out. I am trying to ex­plain why the meaning is always different and you arestill fretting with the letters and words. After all, lettersand words are not just marks on paper; they are signs.The physical marks may be different, but the signs arethe same.

S. I see. We can solve our problem by not talking aboutphysical signs but about physical marks that representsigns?

P. Frankly; Socrates, you are trying my patience. If youwill forget about the marks, we can go on to talk aboutmeaning.

S. You must forgive a slow old man, Psychologus. As yousaid, we are all getting older every minute. But I wasunder the impression that we wcre talking about mean­ing all the time.

P. What do you mean?S. Psychologus, I admit that I'm thinking of a much

simpler kind of meaning than rainbows and heartsleaping up. Those matters are far too complicated fora person like me to describe. They are so complicatedthat r can never quite remember whether they meantthe same to me at two different times. I've a very poormemory, you know, which is why I like philosophy.In a philosopher it can even be an advantage to forgethis old ideas. Now where were we?

P. You said we were talking about meaning all the time.S. Well, I think so~if,meaningis something that is rep­

resented by marks and sounds and the like.P. That's right.

35

t'l

Chapter 2: Meaning and Implication

S. Well, since all those different marks represent signs,I was wondering if what you called the signs and wordsof that little poem aren't meanings just as much as rain­bows and hearts leaping up?

P. Of course not.S. Well then, what name should we give t~ the sort of

thing that is represented by the different physical marksin those different books?

P. I've already said they are signs. You could call themwords or phonemes or whatever you like, so long asyou don't call them meanings. My friend Seispers callsthem "types." The different physical marks are "to­kens," and what they represent is a "type."

S. And a type is not a meaning?P. It is certainly not what I call a meaning.S. Well, let us by no means call it that. But still some­

thing troubles me.P. About meaning-at last?S. Well, about how a type can be the same when the

physical marks that represent it can be so differ-ent.P. What is so strange about that?S. I was wondering how I could think that a type was the

same when each of my experiences of it, my attitudestoward it; -my responses to it are so different. Youknow, whether I am hungry or sleepy, or happy or inpain, whenever I encounter those different tokens I stillthink that they represent the same type.

P. That is precisely what 1 am getting at. The words ofthe poem are always the same, though their meaningis always different.

S. Ah, thank you, Psychologus. You have clarified mythoughts on these matters.

P. Not at all, Socrates. It is a pleasure to talk to a manwho can still continue to learn at such an advanced age.

End of Dialogue

36

L.

B. Reproducibility: Psychologistic Objection:;

Meaning is an affair of consciousness, and the fundamentalcharacteristic of consciousness, as Hume, for all his psycllol­ogism, acutely observed, is that it is always consciousness ofsomething. 7 One of the most brilliant passages in Colerid~'.:::;'s

Biographia Literaria is his use of this insight to attack thF:empirico-psychologistic notion of perception, according t'J

which the thing that one sees when one looks at a table is one';:,perception of a table. How odd, observed Coleridge, when \"::always supposed that we were seeing a table!8 To speak ofperceptions instead of "tables" is precisely the sort of mis­placed sophistication that is found in the psychologistic accountof meaning, according to which what one understands is realiyone's perception of or response to a meaning. But the remar1~­

able fact of consciousness is that the objects of its awarenc~<;

are not the same as the subjective "perceptions," "processes,"or "acts" which are directed toward those objects. My percep­tion of a visible object like Coleridge's table or of a nonvisibicobject like a phoneme can vary greatly from occasion to oc­casion, and yet what I am conscious of is nevertheless the sametable, the same phoncme. This uJliversal ract of consciousnesscannot be explained in ordinary psychologistic terms.v Eitherit must be ignored, or its existence, by some circumlocution,denied.

The. goal-directedness of mental acts, by virtue of whichsomething can remain the same for consciousness even thoughone's perspective, emotion, state of health may vary, is par-

7. See Treatise of Human Nature, Bk. I, Sec. 6: "When 1 entermost intimately into what 1 call myself, 1 always stumble on someparticular perception or other, of heat or cold, light or shade, love orhatred, pain or pleasure. 1 never can catch myself at any time withouta perception, and never can observe anything but the perception."

8. Biographia Literaria, ed. 1. Shawcross (2 vols. London, 1907),1, 179.

9. See Burne, Treatise, Bk. I, Appendix: "If perceptions are dis­tinct existences, they form a whole only by being connected together.But no connexions among distinct existences are ever discoverable byhuman understanding."

If

Chapter 2: Meaning and Implication

ticularly important in a consideration of meaning. This dis­tinction between what is "going on in the mind" on the onehand, and what the mind is averted to on the other, is not,however, a special conception devised for its convenience indefending the self-identity of verbal meanings. It is a character­istic element in all acts of consciousness.

In phenomenology, the philosophical tradition that has mostfully explored the distinction between mental objects andmental acts, this object-directedness of consciousness has beencalled "intentionality"-a word that must be accepted forwant of a better. 10 In the standard phenomenological vocab­ulary the basis for my criticism of the psychologistic concep­tion of meaning would be stated as follows: An unlimited num­ber of different intentional acts can intend (be averted to) thevery same intentional object. Since meaning, like anything elsethat consciousness is averted. to, is an intentional object (thatis, something there for consciousness), and since verbal mean­ing is a meaning like any other, the point can be made morespecific by saying that an unlimited number of different inten­tional acts can intend the same verbal rneaning. This is, of

.course, the crucial point in deciding whether it is possil;Jle toreproduce a verbal meaning. Like any other intentional object,it is in principle reproducible. The psychologistic denial of thisdoes not stand up to- experience.

What led to this denial in the first place was, I think, a con­sideration that really had no connection with the inadequatetenets of the psychologistic position. The kind of psychologismthat prevails among skeptical interpreters usually amounts to aconfusion of verbal meaning with significance-a confusionthat I have already tried to unravel. ll When someone says,"My response to a text is different every time I read it," he iscertainly speaking the truth; he begins to speak falsely when heidentifies his response with the meaning he has construed. Fur­thermore, he is wrong when he identifies his response with

10. For a definition of intentionality see Appendix I, pp. 217-21.11. See Chap. 1, Sec. B, and Chap. 4, Sec. C.

38

B. Reproducibility: Psychologistic Objections

sUbje~tive acts alone. As soon as he makes his own responsean object of consideration, he is concerned with another kindof meaning (i.e. significance) that is potentially as determinateand reproducible as verbal meaning itself. The fact that he candiscuss, remember, describe, and even write about his responseproves this point beyond doubt.

This is not to deny that an interpreter's response-that is,the more or less personal significance he attaches to a verbalmeaning-cannot actually alter the character of the verbal~eaning he construes. Of course this can happen, and it mayIII fact happen very frequently. However, it generally does soprecisely because the interpreter has not troubled to distin­?uish b~tween his response and what he is responding to-anIllustratIOn of the way interpretive theories tend to confirmthemselves. If a reader cannot distinguish between what some­one's text means and what it means to himself, then obviouslyfor such a reader the distinction could have no empirical con­firmation. It is therefore of some practical value to rememberthat neither in fact nor in logic is a verbal meaning the same asany of the countless relational complexes within which it canform a part.

If this distinction is made, there is rio important reason foranyone to insist on the unlikely and untestable hypothesis thatone man's verbal meaning is always necessarily different fromanother's, for the primary cause of the insistence has been aconfusion of verbal meaning with significance. It is true thatthe significance of a text for one person is not altogether thesame as for another, because the men themselves and thereforetheir personal relationships to a particular verbal meaning aredifferent. But this undoubted fact cannot legitimately be ex­tended t~ verbal meaning as well as personal significance. g.r, v'as expenence shows, the same meaning can be intended bydifferent intentional acts of one person at different moments intime, then that is a reasonable warrant for the hypothesis thatthe same meaning can be intended by the different intentionalacts of different personS] And if verbal meaning is, by defini-

39

Chapter 2: Meaning and Implication

tion, meaning that can be shared, then it is reasonable to be­lieve that verbal meaning exists. Obviously, its very existencedepends upon its reproducibility. At the last ditch ~ew would,I think, be so eccentric as to deny the sharability of meaning.To whom and to what purpose would they address their denial?

C. REPRODUCIBILITY: HISTORICISTIC OBJECTIONS

It is one thing to say blankly that we can never "truly" under­stand the texts of a past age; it is quite another thing to venturethe less absolute and no doubt true conception that we some­times cannot possibly~551~ire_.~!~~.~ltu2."~J~iv~A~~S_~~_for understanding an old text. This second stricture obviously'a:p'pii~~'t~m·aii.y-textsfiom~~lturesabout which we know verylittle and also to some from cultures about which we know agreat deal. The absolute form of historical skepticism shouldnot be confused with this healthy consciousness of the limita­tions under which every interpreter sometimes works. Only theabsolute form of radical historicism threatens the enterprise of

.,j re-cognitive interpretation by holding that the meanings of thepast are intrinsically alien to us, that we have no "authentic"access to those meanings and therefore can never "truly" un­derstand them. -

By one of those typical ironies in intellectual history (ironieswhich support Hegel's theory that human thought evolves bynegating itself) it has been a development of historicism itselfwhich in the present day has raised the most persistent objec­tions to the possibility of historical knowledge. Historicismbegan with the belief that all human cultures were immediateto God; that was its root concept in its inaugural years fromHerder to Ranke. Every cultural era was, to ~~~,Ji~Q~!~s

metap~o!, another m~l~~il!!lIj§diy'ip~-=-§YJ1lP£~:rll,Y!,_anc!~~~"melody had its' 0Vo'.? ci,i~ine i.!!.<!iY_~};l_~1~ty.12 Thus historicism

12. J. G. Herder, Sammtliche Werke, ed. B. Suphan (33 'loIs. Berlin,1877-1913),8,314 f.; 18,282 f.

40

- ~ ~-- --=-------=---":._--- -~~----

C. Reproducibility: Historicistic Objections

fIrst insisted that every culture was worth knowing for its ownsake, "as it really was," but with Hegelian and Lovejovian in­evitability this emphasis on the individuality of different cul­tures has now evolved into an emphasis on the unbridgeablegulf between one culture and another. From Dilthey's concep­tion that human consciousness was constituted by its historicalgivens 1il -an idea that was implicit in Herder-it was not avery long step to Heidegger's conception of the temporalityand historicity of human being. The earlier emphasis on in­dividuality which had given signifIcance to the study of othercultures in their own right became, by one or two turns of theHegelian gyre, an emphasis on the impossibility of studyingother cultures in their own right. The past became "ontica1lyalien" to us.

This philosophical form of radical historicism lent intellec­tual respectability to a prevalent and popular form of historicalself-consciousness which had already created an atmosphereof skepticism regarding the genuine knowability of past cul­tures . .By popular historicism I mean the kind of assumptionsunderlying, [or example, all the recurrent magazine articlesthat gravely describe the latest portentous peculiarities of thelatest "younger gcncration," or those assumptions underlyingthe cult of the new and the feeling that one can or cannot thinkor act in a' particular way "in this day and age." The possibleexamples are so numerous and the assumptions so widelyspread and so deeply engrained in the popular mind that suchhistoricism is capable of making itself true. For in the realm ofculture a belief or opinion is as real as an empirical fact and,given enough currency, becomes itself an empirical fact thatmust be reckoned with. Consequently, the popular emphasison the radical differentness of cultural eras-or even on theradical differentness between onc decade and another-has

13. G. Misch and others, eds., Wilhelm Dilt!leys GesammrltrSchriften (8 'loIs. Leipzig and Berlin, 1913-36), 7, 38: "Denn manstosst hier eben an die Geschichtlichkcit des menschlichen Bewusst­seins als eine Grundeigenschaft desselben."

41

Chapter 2: Meaning and Implication

tended to obliterate sensitivity to sameness amid historicalchange and has lent broad credence to the view that we cannot"truly" understand the texts of another age.

This kind of historicism, like the psychologism to which it isintimately related, is not a theory that is capable of empiricalconfirmation or falsification. That its tenets are highly unlikelyto be true I argue in some detail in Appendix II, where Icriticize the only substantial defense of radical historicism inthe field of hermeneutics-that of H. G. Gadamer. Here I shallsimply develop a few brief distinctions that will isolate thegeneral dogma of historical skepticism from the more limitedand reasonable doubts that any interpreter might entertain ina particular case with respect to understanding a particularcomplex of verbal meanings from the past.

In the first place, radical historicism should be distinguishedfrom the popUlar, indeed nearly universal, conviction thatevery age must reinterpret for itself the texts of the past. Thisdoctrine is as much a description of fact as a moral imperative:Every past age has done just that, and every future age willl.10 doubtcontinue to do so. However, it is a mistake to viewthis doctrine as equivalent to the radical historicist dogma thatevery age understands the texts of the past differently, and thatno age truly understa!1ds them as they were, for it is not truethat a "reinterpretation" is the same as a "different under­standing." To think so is to identify an understanding of a textwith the peculiarities and complexities of written interpreta­tion;it is to confuse the subtilitas intellegendi with the sub­tilitas explicandi. This distinction is laid out at greater lengthin Chapter 4, but I have mentioned it briefly here becausefailure to be aware of it would reinforce the plausibility ofradical historicism.I 4

Another distinction that should be drawn is that between thegeneral probability that we can understand a contemporarybetter than a predecessor, and the particular probability that

14. See Chap. 4, Sees. A and B.

42

C. Reproducibility: Historicistic Objections

may obtain in a particular case. It is generally probable that a:"oman will live longer than a man, but this general probabilityIS a useless abstraction when we confront a healthy man offifty and a woman of the same age who has lung cancer. It isaltogether possible, for example, that Lucan was better under­stood by Housman than by many of Lucan's contemporaryreaders, and it is even more probable that Blake is betterunderstood by scholars today than he was understood by anyof his contemporaries. It should be remembered that the lan­guage and assumptions within a culture can be highly variable,so that it might easily be the case that a modern reader couldhave learned the partiCUlar language of a particular authormore intimately than any contemporary who spoke the "same"language.

This last point discloses one of the most vulnerable concep­tions' in radical historicism. The radical historicist is rathersentimentally attached to the belief that only our own culturalentities have "authentic" immediacy for us. That is why wecannot "truly" understand the texts of the past, such "true" ..undcrstandi.ilg being reserved for contemporary texts, and alI­understanding of the past being "abstract" and "constructcd.'3But, in fact, all understanding of cultural entities past orpresent is "constructed." The various languages of a culture(taking "language" in the broadest possible sense) are acquiredthrough learning, and not inborn. Furthermore, since all thevarious languages of a culture are learned by more than oneperson, they can, implicitly, be learned by any person whotakes the trouble to acquire them. And once a person has trulyacquired a language it does not matter how he managed to doso-whether by rote and constant exposure like a three-year­old or by disciplined application and self-conscious design.There is no immediacy in understanding either a contemporaryor a predecessor, and there is no certainty. In all cases, whatwe understand is a construction, and if the construction hap­pens to be unthinking and automatic, it is not necessarily morevital and authentic for that. .

43

n!

Chapter 2: Meaning and Implication

One can make distinctions, present examples, expose mis­conceptions, but one can never prove or disprove the dogmaof radical historicism.l 5 We can never be sure that we have"truly" understood a text from the past any more than we canbe sure we have understood one from our own time. Qenerally,we are more likely to get a contemporary text right, but thisgeneral likelihood does not automatically hold in any particularinstance (where factors of temperament, knowledge, diligence,and luck are decisive), and interpretation is always concernedwith particular texts. But while the position of radical his­toricism is very probably false, one must acknowledge that itsadherents, particularly those of a Heideggerian cast, hold to itstenets as to a religion-and the claims of a religion are ab­solute. Ultimately one simply accepts them or rejects them.

D. DETERMINACY: VERBAL MEANING AND TYPIFICATION

Reproducibility is a quality of verbal meaning that makes in­terpretation possible: if meaning were not reproducible, itcould not be actualized by someone else and therefore couldnot be understood or interpreted. Determinacy, on the otherhand, is a quality of meaning required in order that there besomething to reproduce. Determinacy is a necessary attributeof any· sharable meaning, since an indeterminacy cannot beshared: if a meaning were indeterminate, it would have noboundaries, no self-identity, and therefore could have noidentity with a meaning entertained by someone else. Butdeterminacy does not mean definiteness or precision. Un­doubtedly, most verbal meanings are imprecise and ambigu­ous and to call them such is to acknowledge their determinacy:the; are what they are-namely ambiguous and imprecise­and they are not univocal and precise. This is another way ofsaying that an aI!!!:lXK1.!9H§ meaning has a boundary like anyother verbal me~nhi'g~-and that one of the frontiers on this

15. See Appendix II, pp. 256-58.

44

D. Determinacy: V erbal Meaning and Typi/icatio:

boundary is that between ambiguity and univocality. Som.parts of the boundary might, of course, be thick; that is, theremight at some points be a good many submeanings that b,;.longed equally to the meaning and not to it-borderline mean.­ings. However, such ambiguities WOUld, on another level,simply serve to define the character of the meaning so thai.any overly precise construing of it would constitute a mis­understanding. Determinacy, then, first of all means self-iden­tity. This is the minimum requir,ement for sharability. Withoutit neither communication nor validity in interpretation wouldbe possible.

But by determinacy I also mean something more. Verbalmeaning would be determinate in one sense even if it weremerely a locus of possibilities-as some theorists have con­sidered it. However, this is a kind of determinacy that cannotbe shared in any act of understanding or interpretation. Anarray of possible meanings is no doubt a determinate entityin the sense that it is not an array of actual meanings; thus, it toohas a boundary. But the human mind cannot entertain a pos­sible rneaning; as soon as the .meaning is entertained it is actual."In that case, then," the proponent of such a view 11light argue,"let us consider the text to represent an array of different,actual meanings, corresponding to different actual interpreta­tions." But this escape from the frying pan leads right into theamorphous fire of indeterminacy. Such a conception real1ydenies the self-identity of verbal meaning by suggesting thatthe meaning of the text can be one thing, and also another,different- thing, and also another; and this conception (whichhas nothing to do with the ambiguity of meaning) is simply adenial that the text means anything in particular. I have alreadyshown that such an indeterminate meaning is not sharable.Whatever it may be, it is not verbal meaning nor anything thatcould be validly interpreted.

"Then," says the advocate of rich variousness, "let us bemore precise. What I really mean is that verbal meaning ishistorical or temporal. It is something in particular for a span

45

Chapter 2: Meaning and Implication

of time, but it is something different in a different period oftime." Certainly the proponent of such a view cannot be re­proached with the accusation that he makes verbal meaningindeterminate. On the contrary, he insists on the self-identityof meaning at any moment of time. But, as.I have pointed outin my critique of Gadamer's theory (Appendix II), this remark­able, quantum-leap theory of meaning has no foundation inthe nature of linguistic acts nor does it provide any criterion ofvalidity in interpretation. 16 If a meaning can change its identityand in fact does, then we have no norm for judging whetherwe are encountering the real meaning in a changed form orsome spurious meaning that is pretending to be the one weseek. Once it is admitted that a meaning can change its c?ar­acteristics, then there is no way of finding the true Cinderellaamong all the contenders. There is no dependable glass slipperwe can use as a test, since the old slipper will no longer fit thenew Cinderella. To the interpreter this lack of a stable norma­tive principle is equivalent to the indeterminacy of meaning.As far as his interests go, the meaning could have been definedas indeterminate from the start and his predicament wouldhave been precisely the same.

When, therefore, I say that a verbal meaning is determinateI mean that it is an entity which is self-identical. Furthermore,I also mean that it IS an entity which always remains the samefrom one moment to the next-that it is changeless. Indeed,these criteria were already implied in the requirement thatverbal meaning be reproducible, that it be always the same indifferent acts of construing. Verbal meaning, then, is what it isand not something else, and it is always the same. That is what

I mean by determinacy.A determinate verbal meaning requires a determining will.

Meaning is not made determinate simply by virtue of its beingrepresented by a determinate sequence of words. Obviously,any brief word sequence could represent quite different com-

16. See pp. 249-50.

46

D. Determinacy: Verbal Meaning and Typification

plexes of verbal meaning, and the same is true of long wordsequences, though it is less obvious. If that were not so, com­petent and intelligent speakers of a language would not dis­agree as they do about the meaning of texts. But if a deter­minate word sequence does not in itself necessarily representone, particular, self-identical, unchanging complex of meaning,then the determinacy of its verbal meaning must be accountedfor by some other discriminating force which causes the mean­ing to be this instead of that or that or that, all of which it couldbe. That discriminating force must involve an act of will, sinceunless one particular complex of meaning is willed (no matterhow "rich" and "various" it might be), there would be nodistinction between what an author does mean by a wordsequence and what he could mean by it. Determinacy of verbalmeaning requires an act of will.

It is sometimes said that "meaning is determined by context,"but this is a very loose way of speaking. It is true that the sur­rounding text or the situation in which a problematical wordsequence is found tends to narrow the meaning probabilities forthat particular word sequence; otherwise, interpretation \vouldbe hopeless. And it is a measure of stylistic excellence in anauthor that he should have managed to formulate a decisivecontext for any particular word sequence within his text. Butthis is certainly not to say that context determines verbal mean­ing. At best a context determines the guess of an interpreter(though his construction of the context may be wrong, and hisguess correspondingly so). To speak of context as a deter­minant is to confuse an exigency of interpretation with anauthor's determining acts.l 7 An author's verbal meaning islimited by linguistic possibilities but is determined by hisactualizing and specifying some of those possibilities. Cor­respondingly, the verbal meaning that an interpreter construesis determined by his act of will, limited by those same pos­sibilities. The fact that a particular context has led the inter-

17. On the nature of a context see Chap. 3, Sec. B, pp. 86-88.

47

Chapter 2: Meaning and Implication

preter to a particular choice does not change the fact that thedetermination is a choice, even when it is unthinking andautomatic. Furthermore, a context is something that has itselfbeen determined-first by an author and then, through aconstruction, by an interpreter. It is not something that issimply there without anybody having to make any determina­tions.

While the author's will is a formal requirement for any deter­minate verbal meaning, it is quite evident that will is not thesame as meaning. On the other hand, it is equally evident thatverbal meaning is not the same as the "content" of which anauthor is conscious. That point has already been made inChapter 1.18 An author almost always means more than he isaware of meaning, since he cannot explicitly pay attention toall the aspects of his meaning. Yet I have insisted that meaningis an affair of consciousness. In what sense is a meaning an

. f . ?object of consciousness even when one IS not aware 0 It.Consider the example given in the earlier passage just referredto, in which a speaker admits he meant something he was notaware of meaning. Such an admission is possible because heconceived his meaning as a whole, and on reflection laterperceived that the unattended meaning properly falls withinthat whole. That is, in fact, the only way the speaker's admis­sion could be true..

What kind of whole is it that could contain a meaning eventhough the meaning was not explicitly there? And J:1ow ca.n sucha generous sort of entity still have very stern bhrriers· whichexclude other meanings that the author might actually havebeen attending to, as well as countl~ss others that he was not?Clearly this remarkable characteristic of verbal meaning is thecrucial one to examine.

Suppose I say, in a casual talk with a friend, "Nothingpleases me so much as the Third Symphony of Beethove~."

And my friend asks me, "Does it please you more than a SWIm

18. Sec. E. See also this chapter, Sec. E.

48

D. Determinacy: Verbal Meaning and Typi{icatio,'

in the sea on a hot day?" And I reply, "You take me te,literally. I meant that no work of art pleases me more th:.Beethoven's Third." How was my answer possible? How d;;:Iknow that "a swim in the sea" did not fall under what I meat; '.by "things that please me"? (The hyperbolic use of "nothing ,to stand for "no work of art" is a common sort of linguist';,;extension and Can constitute verbal meaning in any contextwhere it is communicable. My friend could have understoodme. He misunderstands for the sake of the example.) Since Iwas not thinking either of "a swim in the sea" or "Brueghel'sHay Gathering," some principle in my meaning must causeit to exclude the first and include the second. This is possiblebecause I meant a certain type of "thing that pleases me" andwilled all possible members belonging to that type, even thoughvery few of those possible members could have been attendedto by me. Thus, it is possible to will an et cetera without in thelea~t being aware of all the individual members that belongto It. The acceptability of any given candidate applying formembership in the et cetera depends entirely on the type ofvvhole meaning that I willed. That is to say, the acceptability ofa submeaning depends upon the author's notion of the sub..suming type whenever this notion is sharable in the particularlinguistic circumstances.

The definition of verbal meaning given earlier in this chaptercan now be expanded and made more descriptive. I said beforethat verbal meaning is whatever an author wills to convey byhis use of linguistic symbols and which can be so conveyed.Now verbal meaning can be defined more particularly as awilled type which an author expresses by linguistic symbolsand which can be understood by another through those sym­bols. It is essential to emphasize the concept of type since it isonly through this concept that verbal meaning can be (as it is)a determinate object of consciousness and yet transcend (asit does) the actual contents of consciousness.

A type is an entity with two d~cisive characteristics. First, itis an entity that has a boundary by virtue of which something

49

E. DETERMINACY: UNCONSCIOUS AND SYMPTOMATICMEANINGS

E. Determinacy: Unconscious and Symptomatic Meanings

"Buonaparte" in another usage. But that would simply meanthat they are different types as well as, on another level, in­stances of the same type. However, they could never be merelyconcrete instances. The determinacy and sharability of verbalmeaning resides in its being a type. The particular type that itis resides in the author's determining wil1. A verbal meaning isa willed type. The rest of this chapter and most of the nextwill be concerned with the ramifications of this concept andwith its capacity to clarify the nature of verbal meaning andtextual interpretation.

5~

The fact that verbal meaning has to have some kind of bound­ary in order to be communicable and capable of valid interpre­tation does not exclude so-called unconscious meaning. Theonly requisite is that an unconscious meaning, whatever itscharacter, must lie within the boundary that determines theparticular verbal meaning that is being considered. In otherwords, the principle for excluding or accepting unconsciousmeanings is precisely the same as for conscious ones. In manycases it is impossible to be sure whether a meaning was con­scious or unconscious to an author, and in these cases, there­fore, the distinction is irrelevant. However, it is neverthelessserviceable to clarify the concept of unconscious meaning inorder to avoid confusing an author's verbal meaning with .hispersonality, mentality, historicity, and so on, interesting andrelevant as these may be to the legitimate concerns of criticism.

The one negative characteristic common to all varieties ofunconscious meanings is that the author was not aware of them.Obviously, this definition is not very reassuring since there isno limit to what an author may not be aware of. Usually theterm "unconscious meaning" refers to those meanings whichare not attended to by the author but which are nevertheless19. See Appendix III, pp. 266-69.

50

Chapter 2: Meaning and Implication

belongs to it or does not. In this respect it is like a class, thoughit has the advantage of being a more unitary concept: a typecan be entirely represented in a single instance, while a classis usually thought of as an array of instances. The sec~nd de­cisive characteristic of a type is that it can always be repre­sented by more than one instance. When we say that two in­stances are of the same type, we perceive common (identical)traits in the instances and allot these common traits to the type.Thus a type is an entity that has a boundary by virtue of whichsomething belongs to it or does not, and it is also an entitywhich can be represented by different instances or differentcontents of consciousness. It follows that a verbal meaning isalways a type since otherwise it could not be sharable: If itlacked a boundary, there would be nothing in particular toshare; and if a given instance could not be accepted or rejectedas an instance of the meaning (the representational characterof a type), the interpreter would have no way of knowing whatthe boundary was. In order that a meaning be determinate foranother it must be a type. For this reason, verbal meanings, i.e.shared meanings, are always types and can never relinquishtheir type character. 19

Thus verbal meaning can never be limited to a unique,concrete content. It can~ of course, refer to unique entities, butonly by means that transcend unique entities, and this tran­scendence always has the character of a typification. This is soeven when a verbal meaning has reference to something thatis obviously unique, like "the death of Buonaparte." "Death,""the," and "of" all retain their type character even though theircombination might effect a particular new type. The same istrue of "Buonaparte," for a name is a type, and the particularname "Buonaparte" could not relinquish its type characterwithout thereby ceasing to be a name, in which case it wouldbe incomprehensible and unsharable. No doubt this particularname in a particular use would not have a meaning identical to

I .. \i:.'

E. Determinacy: Unconscious and Symptomatic Meanin;;s

tary indication of a disease. 2o Symptomatic meanings may beof immense interest, but they should not be confused withverbal meanings, because verbal meaning thereby loses itsdeterminacy. There is no limit to the different things a text canbe symptomatic of, and there is no intrinsic reason to limitsymptomatic meanings simply to those inhabiting the author'ssubconscious mind.

On the other hand, it would be a mistake to draw the linebetween a sign and a sympt~m by a simple and crude dis­crimination that ignores the variability and latitude of verbalmeanings. If, for example, a husband comes home to his wife,sighs deeply, and says, ''I'm very tired tonight," his verbalmeaning might contain, in addition to information about hisphysical state, a plea for sympathy and praise. Even if thisplea were largely unconscious it might still be part of the verbalmeaning if the conventions established by habitual usage be­tween the husband and wife made it possible for the words''I'm very tired tonight" to convey such an implicit plea. Partof the convention might be that the phrase must be utteredwith a shake of the head and a deep sigh, that it must not bestated with reference to particular plans for the evening, but

J only a-propos des bottes, and that it must be said only when thehusband is known to have been working hard. Once theseconventions have become established (and all verbal meaningrequires analogous generic conventions, as I point out in thenext chapter), then it is not necessary that the husband alwaysattend explicitly to all the implications of his utterance, thoughhe must consciously will a particular type of meaning in orderfor the meaning to exist at all. Verbal meanings of this sort arelike icebergs: the larger part may be submerged, but the sub­merged part has to be connected with the part that is exposed.

The iceberg metaphor presents the image of a visible shapeconnected to a larger invisible shape below the level of con­scious awareness. Even though the visible mass is the smaller

20. See Charles Bally, "Qu'est-ce-qu'un signeT' Journal de Psycho­logie normafe et pathologique, 36 (1939), 161~74.

5352

Chapter 2: Meaning and Implication

present in another region of his mind-a lower region as itwere, which is generally called the subconscious. The term isnormally restricted still further to those meanings in the au­thor's subconscious mind which are indicated by characteristicsof his text. While this last very sensible limitation approachesthe criteria for verbal meaning as defined in this ~hapter, itdisregards one crucial element of the definition, the element

of will.While it is possible to will a great many things of which one

is not directly aware (for example, the continuation of an etcetera), it is not possible to will something against one's will.That is a verbal contradiction which discloses a contradictionin fact. Will can extend into unknown and unnoticed regions asfar as it likes, but it cannot relinquish its connection with thataspect of itself which is conscious. For will involves not merelychoices and goals, but voluntary choices and goals, and againour habits of language remind us of the conscious element inwill. A "tendency" or "impulse" that is totally subconscious,that has no strands tying it directly to a conscious impulse, isnot willed in the ordinary sense of the term, nor in the sense1 allot to the word. Such an impnlse would be, precisely, in­voluntary. And even if such an involuntary impulse were dis­closed in speech, that would not in itself make it a constituentof verbal meaning.

One obvious ex~mple is stuttering. The fact that a personstutters when he speaks certain words may indicate a great dealabout him, but these indications are not part of his· verbalmeaning. They are, rather, involuntary accompaniments tomeaning, that is, symptoms of meaning, not linguistic signsrepresenting meaning. The difference between a sign and asymptom consists precisely in this: a sign is voluntary (ar­bitrary) and conventional, a symptom involuntary and inde­pendent of convention. A linguistic sign is able to representa range of verbal meanings precisely by virtue of its arbitrarycharacter, while a linguistic symptom is a nonarbitrary indica­tion of something else, just as a fever is a symptom or involun-

,,,:;,

.;A,

.1..

;;

.(

;:I,i

\1rI~ .

,.'J

Chapter 2: Meaning and Implication

part, it determines, from the standpoint of anyone examiningthe iceberg, what belongs to the iceberg as a whole and whatdoes not belong. Any part of the whole that is not continuouswith the mass above the surface cannot be part of the iceberg.If there is something down below which is separate, and dis­continuous, then it must either be independent or belong tosomething else. Physical analogies are dangerous, but in thiscase the analogy holds. The self-identity of a verbal meaningdepends on a coherence that is at least partly analogous tophysical continuity. If a text has traits that point to subcon­scious meanings (or even conscious ones), these belong to theverbal meaning of the text only if they are coherent with theconsciously willed type which defines the meaning as a whole.If such meanings are noncoherent with the willed type, thenthey do not belong to verbal meaning which is by definitionwilled. As soon as unwilled meaning is admitted, then anythingunder the surface of the vast sea could be considered part ofthe iceberg, and verbal meaning would have no determinacy.

But can the distinction between a sign and a symptom bemade in practice? E1.o\-v is one to judge whether a particularmeaniug is coherent or noncoherent with the willed type? Theprinciple of coherence is precisely the same as the principle ofa boundary. Whatever is continuous with the visible part of aniceberg lies inside its. boundaries, and whatever lies withinthese falls under the criterion of continuity. The two conceptsare codefining, and I have already shown that the bou.ndaryprinciple depends on the concept of a type. Any meaning thathas the trait or traits by which a type is defined belongs to thattype, and any meaning which lacks these traits does not belong.The principle of continuity is that of membership in a type. Inother words, as I stated at the beginning of this section, theprinciple for accepting or rejecting unconscious meanings isprecisely the same as for conscious ones.

The adequacy of the conception can be illustrated by theexample of lying. Does the verbal meaning of a lie consist inthe meaning that a speaker wills to convey, or does it also carrythe additional meaning that what is willed is deliberately false?

54

E. Determinacy: Unconscious and Symptomatic Meanings

If a lie did carry this additional meaning, which is antagonisticto the usual purpose of lying, then on most occasions therewould be no point in telling a lie. In other words, if part of theverbal meaning of a lie were that it is false, then there wouldreally be no such thing as a lie, since one part of the meaningwould rectify the falsity of the other part. We do not say thatsomeone has misunderstood a lie when he is taken in by it.Re has understood it only too well; the liar's verbal meaninghas been successfully communicated.

But consider the case of the unsuccessful lie or, shall we say,the stylistically inept lie. A boy plays hookey. His mother askshim later to tell her what happened that day at school. The boyblushes' deeply and hesitates: "Oh, er, just the usual thing. Ihad arithmetic and er geography. Oh, no, that's wrong; wedidn't have geography today. It was English and er," and thisis broken off with a gesture of uncertainty. We might supposethat the story had been insufficiently rehearsed or, better, thatsubconsciously the boy did not want to lie. But whatever con­clusion we might draw, the fact rcmains that the boy did lie.His verbal rncaning was raIse. His stylistic incompetcllce wasnot part of his verbal meaning but was symptomatic of hisconscious or subconscious unwillingness to lie.

I choose this extreme example because borderline cases areoften the most informative ones. If verbal meaning is deter­mined by will and if, as in this case, a text seems to discloseantithetical impulses, how can the principle of a willed typeprovide a criterion of coherence? Are there not two disjunctivewilled types, and therefore is the meaning not much morecomplex than the simple conceptual model would suggest? Ithink the proper answer is that the conceptual model showsprecisely how to clarify such complexities. As long as the boycontinued to lie, the willed type represented by his words in­cluded the meaning that he had been at school and excludedthe meaning that he had not been at school. The truth-tellingimpulse that he might also have willed lay outside his verbalmeaning because it could not be communicated by his words.If he suddenly broke off and confessed, then the meaning of

55

Chapter 2: Meaning and Implication

his second statement would contradict that of the first, and therpeaning of the second statement would be a contrary willedtype. Thus his ineptitude may have been symptomatic of adivided will, but his verbal meaning, limited as it was by thelinguistic signs he employed, was a unity_ . I

This insistence on the unity of verbal meaning does not ex-clude the notion of a divided will when it is expressed as a signrather than as a symptom. For example, if the boy had said,"Well, er, maybe I was at school today," then his unwillingnesseither to lie or to tell the truth would have been expressedverbally in an ambiguous willed type, and his verbal meaningwould be ambiguous, since the word "maybe" functions as averbal sign rather than a symptomatic accompaniment. Fur­thermore, since the ambiguity of the boy's will is now directlypart of his verbal meaning, his halting hesitancies of speechcease to be merely symptomatic accompaniments and becomestylistic reinforcements of meaning. The reason that the hesi­tancies should no longer be considered "involuntary" symp­toms lying outside the boundary of verbal meaning is that theyare now cxnressions of a will that lies within the verbal willedtytlC instca~t of an accompanying i.mpulse that lies outside its

boundary.However, it would be very foolish to say that symptomatic,

involuntary meanings are not a proper and legitimate concernof criticism. In fact, they are one of the most interesting sub­jects of critical inquiry. Obviously the most profitable thing toknow about a lie is that it is a lie-an act of judgment thatentirely depends on distinguishing a man's verbal meaningfrom thg symptoms and facts that may betray him. WhenBlake said that Milton wrote in fetters when he spoke of angelsand at liberty when he spoke of devils, because he was of thedevil's party without knowing it, his entirely legitimate criticalcomment was not necessarily a comment on the verbal meaningof Paradise Lost. 'll It was primarily a symptomatic inference.

21. The Marriage of Heaven and Hell, PI. 6.

56

F. Determinacy: Meaning and Subject Matter

Of course, it is a far more interesting critical observation thana mere interpretation of verbal meaning usually is-interestingbecause it is a comment on Milton, and on poets, and becauseit implicitly asserts the superiority of Books I and II overBook III of Paradise Lost, a kind of critical judgment that noone would want to exorcise from literary criticism.

Thus, when I make the point that symptomatic inferencesare not interpretations of verbal meaning, my purpose is not~o s~~gest that such inferenecs are in some way impure orIllegItImate, but to clarify the distinction made in Chapter 1betwe.en ~eaning and significance. Symptomatic, involuntarymeanmg IS part of a text's significance, just as its value or itspresent relevance is. But significance is the proper object ofcriticism, not of interpretation, whose exclusive object is verbalmeaning. It is a charter of freedom to the critic, not an inhibi­tion, to insist on this distinction, for the liberty of the criticto describe the countless dimensions of a text's significance isclosely dependent on his not being constricted by a confusionbetween significance and meaning. No responsible critic wantsto pervert and falsify the lncc1l1ing of a text, yel ,"1. the S',1rne timehe dOES nell want to be inhibited from pursuing v!llatcvcr SeCJT1S

most valuable and useful. 22 If he recognizes that verbal mean­ing is determinate, whereas significance and the possibilitiesof legitimate criticism are boundless, he will have overcome aconfusion that has, ironically, inhibited critical freedom. Atthe same time, he will not dismiss lightly the modest, and in theold-fashioned sense, philological effort to find out what anauthor meant-the only proper foundation of criticism.

F. DETERMINACY: MEANING AND SUBJECT MATTER

When discussing Kant's claim to understand Plato better thanPlato himself, I observed that Kant failed to distinguish be­tween Plato's meaning and the subject matter to which that

22. See Chap. 4, Sec. E.

57

Chapter 2: Meaning and Implication

meaning referred. This apparently simple distinction is, how­ever, far from easy to grasp, and if it eluded Kant, it is onlyfair to confess that it quite thoroughly eluded me in myprevious essay on hermeneutic theory (Appendix I). It is alsoa distinction that Hussed failed to observe in what is, nonethe­less, the most detailed, penetrating, and convincing account ofmeaning that 1 am acquainted with (Logische Untersuchungen,Part II). Probably the first methodological, though not totallysatisfactory, approximation of the distinction was made byDe Morgan in his brilliant essay "On the Structure of theSyllogism."2'l In De Morgan's influential terminology the dis­tinction was stated as one between the universe as a whole anda particular "universe of discourse." De Morgan's vocabularyis in this context less serviceable than his ideas, and I havefound it more useful in describing the determinacy of verbalmeaning to define the distinction as one between meaningand subject matter.

The distinction arises from the observable fact that not alluses of a word like "tree" carry the same implications. If some­one heard the word "tree" spoken by a child, a woodsman, abotanist, or a poet, he would very reasonably guess that in eachinstance the word probably carried different implications.Specifically, he might infer that the botanist implied not onlythe part of the tree that is above ground, but the root systemas well. A child, on the other hand, though he could be awarethat a tree has roots, might mean simply the part of the t~ee

that is visible. Yet it is a fact about trees that they have roots.Does this mean that roots· are implied willy-nilly when some­body uses the word "tree"? Apparently not, since people doentertain and communicate implications that are inadequate orfaulty. If the implications of a verbal meaning were invariably

23. Augustus De Morgan, "On the Structure of the Syllogism, andon the Application of the Theory of Probabilities to Quest~ons of Argu­ment and Authority," Cambridge Philosophical TransactIOns (Nov. 9,1846). See also F. Rossi-Landi, Significato, communicazione e parlarecomune? (Padua, 1961), PP' 249-61.

58

F. Determinacy: Meaning and Subject Matter

determined by the "objective" character of what it refers to,then nobody could ever communicate a conceptual mistake!There is, therefore, a distinction between meaning and subjectmatter.

To define the distinction is, however, no simple task sincesubject matter is a concept that apparently makes absoluteepistemological claims. There is, on the one hand, what some­one implies by "tree" and, on the other, what "tree" in factreally implies. But who is to say what "tree" really implies?To assume that there is some independent and universalground of implication that transcends and controls what anyindividual might mean by "tree" is to fall into the fallacy ofthe public consensus, under which a use of the word wouldhave the same implications to all, regardless of the author'smeaning. 1 shall not repeat the arguments of Chapter 1 whichdeny the existence of such public unanimity but, instead, shallconsider the quite relative character of subject matter as adiscriminating concept.

When someone's meaning is incomplete or false, we areable to say that it is inac1cql1atc to its SllbjcCl matter only jf ViC

have or believe that we have a more cOlnpJcte ;:md t:ruer con­ception of the subject matter than the author has. But supposewe in turn express our superior conception of the subjectmatter and are judged by a further critic who believes he has astill truer or broader conception than our own. He in turn willsay that our meaning is inadequate, and he will do so on thebasis of a still different conception of the subject matter. Now,in each case the judgment of the two critics might be correct.The first critic might have a conception that is truly moreadequate than the original author, and the second critic mighthave a conception more adequate than the first. On the otherhand, one or both of the critics might be wrong. Obviously,the notion of subject matter is in practice entirely relative to theknowledge or presumed knowledge of the critic.

Thus, while it may be the case that a critic has once and forall reached a totally adequate conception of a subject matter,

59

-~o;,

Chapter 2: Meaning and Implication

it is also true that this is not always so, and in practice subjectmatter is a variable conception. It would be highly presumptiveof any critic to claim that he had attained absolute knowledge,although he might very reasonably claim a broader knowledgethan the author's. It follows that, in any particular instance ofcriticism, subject matter is an ideal pole of knowledge which isin fact represented by the present conception of the critic. Tosay that meaning may be different from subject matter is to saythat an author's conception of something may be different froma critic's--which is a self-evident proposition.

But this reduction of the distinction to something self­evident does not altogether resolve the very real problem ofdetermining the difference, if any, between meaning and subjectmatter in a particular case. If we believe that any author orany person would agree that a tree has roots, is it not reason­able to assume that roots are implied by the word "tree" inany usage? The author might not have considered this neces­sary implication, but on reflection he would surely agree thathe had to imply roots when he said "tree." This argument is,huwever, mis1cading. It may be true that any reasonable manmight be brought to admit that be should have implied "roots"when he said "tree," and a persuasive critic might even con­vince him that his meaning did in fact carry that implication.But there is a distinction between what an author admits heshould have meant in order to embrace a true conception of atree, and what he might actually have meant.

Furthermore, if the critic's conception of a subject matter ismade the ground for determining the implications of an ut­terance, then it also becomes the ground for determining theirinterrelationships and relative emphases. But a subject matteris surely neutral with respect to these things. If the meaning"roots" is implied by "tree," that still does not indicate whether"roots" is a vague or highly precise meaning, whether thenutritional function of the roots is implied, whether the rootshave a central emphasis or trail off into a dim penumbra ofmeaning. None of this could be answered simply with reference

60

G. Determinacy: Meaning and Implication

to subject matter, and consequently subject matter cannotdetermine implications. With respect to a subject matter theimplications of an utterance remain indeterminate just as theydo with respect to a putative public consensus. The properground for determining implications is now to be considered.

G. DETERMINACY: MEANING AND IMPLICAnON

Most of the practical problems of interpretation are problemsof implication. There are, of course, a good many instanceswhere the most primitive and "literal" meanings of a text cancome under dispute, but these are far rarer than controversieswhich turn on the "unsaid" meanings of a text. In the looseterminology of some literary critics such meanings have beencalled "connotations"-that is, implications "meant-with" themanifest or "denotative" content of a text. This use of "denota­tion" and "connotation" is, of course, at variance with theiruse in logic, and I have abandoned the words altogether be­cause they have lost their precision anel because there is notand could not be a universally applicable distinction betweenprimary or manifest, and secondary or nonmanifest meanings.No meaning represented by a verbal sign is manifest; all mean­ings must be construed, and what is "manifest" in a particularconstruction may not even have been directly noticed by theauthor. (That is why I placed the words "literal" and "unsaid"in quotation marks above.) Of course, some meanings arenecessarily dependent on prior or primary meanings, and con­sequently the words "denotation" and "connotation" docorrespond to a distinction whose application to a particulartext everyone might be in agreement about. However, for thepurpose of adequate theoretical description, it is more usefulto find terms that have both precision and generality. I thinkthe commonly used term "implication" has both qualities. Tosay that a particular meaning is implied by an utterance is notto insist that it is always "unsaid" or "secondary," but only

61

Chapter 2: Meaning and Implication

that it is a component within a larger whole. The distinction isbetween a submeaning of an utterance and the whole array ofsubmeanings that it carries. This array, along with the prin­ciples for generating it, I call the "meaning" of the utterance,and any submeaning belonging to the array ! call an "im-plication." .

Few would deny that the crucial problem in the theory andpractice of interpretation is to distinguish between possibleimplications that do belong to the meaning of a text and thosethat do not belong.24 I have argued that if such a principle ofdeterminacy did not exist (a principle under which we acceptor reject possible implications) communication and interpre­tation would be impossible. The determinacy of verbal mean­ing is entirely dependent on the determinacy of implications­that is, on the existence of a principle for including or excludingthem. Undoubtedly, the most important preliminary principleof discrimination is that which distinguishes verbal meaningfrom significance. That distinction, widely overlooked andalmost entirely unpublicized since the time of Boeckh, is worthrecapitulating before turning to the general problem of im­plication.

If, as I have argued, verbal meaning necessarily has the char­acter of a willed type. that can be conveyed through linguisticsigns, then significance would be any meaning which has arelation to the verbal meaning so defined-no matter howneutral, descriptive, or tame the related meaning. might be.Thus if it is said that Gibbon's comments on superstition re­flect ~he common attitudes of his own time, that would pointout a meaning of Gibbon's work to historical generalities, butnot a meaning in the work itself. The difference between these

24. Classic examples of the problem are found in William Empson,Seven Types of Ambiguity (3d ed. New York, 1955), which demon­strates on almost every page what happens to interpretation when atext is self-consciously conceived to be a "piece of language," and theproblem of validity is ignored. Anyone desiring further concrete ex­amples of the issues probed in this book is advised to consult Empson.

62

G. Determinacy: Meaning and Implication

tiny prepositions is highly important and too often ignored. Sig­nificance is always "meaning-to," never "meaning-in." Signifi­cance always entails a relationship between what is in a man'sverbal meaning and what is outside it, even when that relation­ship pertains to the author himself or to his subject matter.If Milton really was of the deviI's party without knowing it, thatwould be part of the meaning of Paradise Lost to Milton'spersonality, part of the work's significance, and no doubt suchobservations do call attention to characteristics of meaning inParadise Lost. (Criticism and interpretation are not, as I pointout in a subsequent chapter, autonomous.) If such instances ofsignificance are not distinguished from instances of meaningthe result is bound to be a now familiar state of confusion, forthere is literally no limit to the significance of the shortest andmost banal text. Not only can its verbal meaning be relatedto all conceivable states of affairs-historical, linguistic, psy­chological, physical, metaphysical, personal, familial, national-but it can also be related at different times to changing con­ditions in all conceivable states of affairs. Not that such ex­ercises would be very often useful or interesting, but theycould be performed, and that which is interesting or usefulto somebody varies considerably with different men and. dif­ferent times. When, in the preceding two sections, I distin­guished verbal meaning both from symptomatic meaning andfrom subject matter, I simply selected for examination the twokinds of significance that have been most often confused withverbal meaning. However, there are innumerable varieties ofsignificance beyond these, and plenty of breathing space for allconceivable exercises of criticism so long as it emancipates it­self from the inhibitions of a state of confusion.

While significance is by nature limitless, the crucial featureof implication is that it is not, and the nature of its limits isindicated by the useful, though not completely adequate, meta­phor of its etymological derivation. To be "folded in" is to beinside, ready to be folded out or explicated. The metaphor isnot quite adequate because it suggests that an implication is

63

Chapter 2: Meaning and Implication

always hidden, lurking behind or between the folds of moreobvious or primary meanings. This is very frequently the case,of course, but not always, since, as I have indicated, it is notalways possible to distinguish what is primary or obvious fromwhat is secondary and hidden. Nevertheless, the metaphor isuseful insofar as it suggests that implications lie within themeaning as a whole and are circumscribed by some kind ofboundary which delimits that meaning. Thus the etymologicalmetaphor suggests a more general and, I think, quite indispen­sable conceptual model-that of part and whole. An implica­tion belongs within a verbal meaning as a part belongs to awhole.

A merely spatial conception of the part-whole relationshipis inadequate, however, because it suggests an articulatedphysical object whose parts have the same physical characteras the whole which they constitute. The peculiarity of a wholemeaning is that it retains its integrity and completeness even ifall its implications have not been articulated. In other words,the whole meaning is not simply an array of parts but is also aprinciple for generating "parts," a principle by virtue of whichthe meaning is somehow complete or whole even though theactual job of generating all the parts remains incomplete. Whatis this remarkable principle? I have suggested that it is theprinciple which characterizes a type. 2 [; The special potency ofa type is precisely the same as the generative potency possessedby a meaning. A type stands independent and complete, yet atthe same time it contains a principle by virtue of which it ispossible to judge whether any conceivable entity belongs to orembodies the type. This type principle requires elaboration.

A type is an entity that can be embodied in or represented bymore than one instance. Anything that is unique cannot, withrespect to those aspects which are unique, be a type. Preciselybecause a type can be embodied in more than one instance, ithas the apparently magical potency of containing and generat-

25. Chap. 2, Sec. D.

64

G. Determinacy: Meaning and Implication

jng parts of itself which it does not explicitly contain. For ex­ample, if we consider a very simple type such as a right triangle,we can say that the type contains the implication stated in thePythagorean theorem. (For the sake of simplicity in exposition,I am assuming that the type is in this case equivalent to subjectmatter, though it is perfectly possible to have a willed type ofright triangle which sharply excludes some of its geometricalproperties.) But why is it that the type "right triangle" containsthe implication, the square of the hypotenuse equals thesummed squares of the other two sides? If one answers, "Be­cause that is the nature of a right triangle," one simply begs thequestion. If one answers, "Because part of the meaning of aright triangle is the Pythagorean theorem," that would be moredescriptive, but it would not explain how "right triangle" cancontain "Pythagorean theorem," particularly if one did not ex­plicitly attend to the theorem when one intended the type. "Butthe theorem applies to all right triangles so it must apply here."This begins to be more illuminating, though we may stillwonder how one meaning "contains" the other. "Since I havelearned, thanks to Pythagoras, that his theorem applies to allright triangles, aDd since almost everybody else has learnedthis too, it is possible to mean 'Pythagorean theorem' as part ofwhat I mean when I say 'right triangle.' If nobody had everheard of the theorem it would not be possible to have it as partof my verbal meaning. Not only does the theorem apply to allmembers of the type, making it a characteristic that belongs tothe type, but it is also something that is known by others tobelong. Because of their knowledge, the theorem is containedin the meaning 'right triangle.' They are able to fill out the im­plications because they are familiar with the type. If they werenot familiar with it they could 110t do so, and I could not conveythe implication."

We have finally managed to arrive at a satisfactory explana­tion. Since a type is something that can be embodied in morethan one instance, it is something whose determining char­acteristics are common to all instances of the type. Further-

65

nmrmcmx:mmr N

-- iEWf -=:wwrrrrZ57zzserE7ZGF" r wnw" I'fl fl'f5fiJ~

Chapter 2: Meaning and Implication

more, since the type can be represented in more than oneinstance, it can be shared or known by more than one person.When another person has learned the characteristics of thetype, he can "generate" those characteristics without theirbeing given to him explicitly. It is sufficient merely to give hima decisive clue as to the particular type that is meant.

An implication belongs to a meaning as a trait belongs to atype. For an implication to belong to verbal meaning, it isnecessary that the type be shared, since otherwise the inter­preter could not know how to generate implications; he wouldnot know which traits belonged to the type and which did not.And there is only one way the interpreter can know the char­acteristics of the type; he must learn them. (For these char­acteristics are not usually "syncategorematic" or absolutelynecessary comeanings like color and extension. Even thePythagorean theorem is a learned characteristic of a righttriangle, no matter how "necessary" it may seem once it islearned.) Implications are derived from a shared type thathas been learned, and therefore the generation of implicationsdepends on the interpreter's previous experience of the sharedtype. 'The principle for generating implications is, ultimatelyand in the broadest sense, a learned convention.

The reader will notice that I have deliberately made a smallalteration in my description of verbal meaning. Instead of call­ing it a "willed type," I have used the expression "shared type."In doing so I have shifted emphasis from the type willed by theauthor to a type experience that is common to, author andreader. This is the other side of the coin. If verbal meaning is awilled type that can be conveyed through linguistic signs, itfollows that the possibility of conveying the willed type dependson the interpreter's prior experience of the willed type. Other­wise, the interpreter could not generate implications; he wouldnot know which implications belonged to the meaning andwhich did not. The willed type must be a shared type in orderfor communication to occur. This is another way of sayingthat the willed type has to fall within known conventions in

66

G. Determinacy: Meaning and Implication

order to be shared-an eXigency that was implicit from thestart in the concept of sharability.

My emphasis in this chapter has been on the author's Will,because my central topic has been the determinacy of verbalmeaning, and authorial will is a formal requirement for deter­minacy. Of equal importance is the sharability of verbal mean­ing, and for this the necessary requirement is the existence ofshared conventions. Verbal meaning is both a willed type anda shared type. This second characteristic is the main subject ofmy next chapter.

67

PROBLEMS AND PRINCIPLESOF VALIDATION

The only proper attitude is to look upon a successfulinterpretation, a correct understanding, as atriumph against odds.

I. A. Richards

A. THE SELF-CONFIRMABILITY OF INTERPRETATIONS

The activity of interpretation can lay claim to intellectual re­spectability only if its results can lay claim to validity. ~n theother hand, its claims need to be moderated to Slllt thepeCUliarities and dilliculties attending the interpre~ivc ~nte:­

prise. Aristotle made the appropriate remark on tlus pom! Inhis Ethics, where he observed that no conclusion should ar­rogate to itself a greater certainty or precision than its subjectmatter warrants. In this section I shall describe a fundamentaldifficulty of interpretation which hinders any neat for~ulati~mof correct methodology and must sober any self-conv1l1ced 111­

terpreter of a tcxt. The fact that certainty is always unattain­able is a limitation which interpretation shares with many otherdisciplines. The special problem of interpretation is that itvery often appears to be necessary and inevitable when in factit never is. This appearance of inevitability is a phantasm raisedby the circularity of the interpretive process.

The belief that written language carries its own indubitableforce has a lineage as ancient as the primitive belief in themagical properties of words. But a nearer so~rce for theendemic (and now epidemic) belief in the semantic autonomy

164

A. The Self-Con{irmability of Interpretations

of language is the fact that interpretation very often induces aprofound sense of conviction. The interpreter is convinced thatthe meanings he understands are inevitable, and this timewornexperience (quite aside from any of our pe«uliarly modernproclivities) has always lent credibility to the idea that mean­ings are directly given by words. When an interpreter maintainshis unruffled certainty in the face of contrary opinions, we mayassume that he has been trapped in the hermeneutic circle andhas fallen victim to the self-confirmability of interpretations.

There lurks a partial, but nonetheless helpful, analogy to theself-confirmability of interpretations in the process of decipher­ing totally unknown sign systems. The memory of Ventris'achievement in this field is still fresh, yet for all its compellingbrilliance, Ventris' decipherment of Linear B was not at firstuniversally accepted. Some scholars very justly objected thatsuch a decipherment had the property of confirming itselfbecause its internal consistency was guaranteed in advance.The decoded elements had been used to construct the verySYStCl11 which ,gavc rise to the decodcd elements. The text Ull-.

failingly confirmed thc theory because there was nothing in thetext which was not sponsored by the theory in the first place:from a mute array of inscrutable signs the only meanings to begleaned were those which were sponsored by the theory theypurported to confirm. Ventris was able to meet this objectionconvincingly only after his decipherment had been further con­firmed by newly discovered texts that had played no part in theconstruction of his hypothesis.

The circularity of such a decipherment, while only partiallyanalogous to the circularity of interpretation, does serve toremind us that a luute sign system must be construed before itfurnishes confirmation of an interpretation. Furthermore, themanner in which the signs are construed is partly predeter­mined by the interpretation itself. When interpretcrs construetexts differently, the data they use to support their construc­tions are to some degree sponsored by those constructions. Sowe confront a very slippery sort of entity when we read a text.

165

167

A. The Self-Confirmability of Interpretations

tween i.nte~pretations is ~ot solved simply by the interpreter'sdetermmatIOn to entertalll alternative hypotheses about h"~ext-tbough that is the necessary precondition for objecti:~Judgment. The interpreter faces the much more difficult prob­~em of comparing hypotheses which are in some respectsIDco~mensurab~e: when a text is construed under differentgene.rIc conceptIOns, some of the data generated by one con­ceph~n will be different from those generated by the other.. ThIS tendency of interpretations to be self-contained andlllcommensurable is,. I ~el.ieve, the principal handicap that willa~ways plague .the dISCIplIne of interpretation. Interpretationshave a propen~Ity P~pe observed in eighteenth-century watches-none ?oes JUs~ alIke, yet each interpreter believes his OWn.ParadOXIcally thIS very proliferation of opinions accounts forunwarrant~d.optimism on the one side and equally unwar­ranted cyn~clsm on the other. The optimist assumes that somany convlll~ed and competent readers cannot be wrong, andh~ therefore VIews their divergences not as .representing genuine~ls~greementsbut as ~e~~c~ing different aspects and potentiali­tIes. of the t~xt: .In C[J:1CIZlI1g this conception, I have alreadyobserved that (hffcrent mlerpretatiolls can indeed be reconciled~ot because .they are complementary but because they somc~tImes take dIfferent paths toward the same generic meaning.]~o~ever, I ~lso observ.ed that sometimes the generic meaningsImplIed. by lllterp.retahons are disparate. To dream that allexp~rt ~nterpretahons a.r~ ultimately members of one happyfamIly IS t<:> a.bandon cnhcal thinking altogether.

The op~Im.Ist.does, in one respect, push closer to the truththan the ll~vlll~l~le cynic who disbelieves all interpretationsequally. HIS wlihngness to adjust and reconcile in order todemo~strate the "area of agreement" shared by different inter­pretah~ns at least avoids the futility of controversy over merelyverbal Issue~ an~ dispels merely apparent disagreement whereno substantral dIvergency exists. But the optimist also glosses

1. See Chap. 4, Sec. A.

?W''Ml'FitDZ

rIi\

166

Chapter 5: Problems and Principles of Validation

The word patterns and stylistic effects which support one in­terpretation can become different patterns and effects under adisparate interpretation. The same text can sponsor quite dif­ferent data (though some of the data will remain constant), andeach set of data will very powerfully support the interpretivetheory which sponsored it in the first place.

I have given one convenient example of this incestuous rela-tionship in my comment on Donne's "Valediction ForbiddingMourning." The words of the text take on a consistent patternof meanings when we suppose that they are spoken by a dyingman but a quite different pattern under a different hypothesis.The disagreements of experts may be harder to resolve thanthis, but they usually follow the same pattern. Every inter­preter labors under the handicap of an inevitable circularity:all his internal evidence tends to support his hypothesis becausemuch of it was constituted by his hypothesis. This is anotherdescription of the relationship between an intrinsic genre andthe implications which it generates. An interpretive hypothesis--that is, a guess about genre-tends to be a self-corrl1rming

hypothesis.Thus, the distressing unwillingness of many interpreters to

relinquish their sense of cer!ainty is the result not of nativeclosed-mindedness but of imprisonment in a hermeneutic circle.Literary and biblical interpreters are not by nature morewillful and un-self-critical than other men. On the contrary,they very often listen patiently to contrary opinions, and aftercareful consideration, they often decide that the contrary hypo­theses "do not correspond to the text." And of course they areright. The meanings they reject could not possibly arise excepton the basis of a quite alien conception of the text. It is verydifficult to dislodge or relinquish one's own genre idea, sincethat idea seems so totally adequate to the text. After all, sincethe text is largely constituted by the hypothesis, how could thehypothesis fail to seem inevitable and certain?

This circular entrapment is not, unfortunately, merely apsychological difficulty. The problem of correctly judging be-

Chapter 5: Problems and Principles of Validation

over disagreement where it does exist and thereby avoids theresponsibility of rational choice. The cynic, on the other hand,quite rightly perceives that disagreements are sometimes finaland irreconcilable. He observes that one interpreter rarely ifever persuades another, because each feels as convinced of ,hisown view as does the cynic himself. He therefore concludesthat the interpreter's sense of conviction cannot be objectivelybased but must arise from the peculiar constitution of theinterpreter himself-his historicity, psychology, personality,and so on. Ultimately, the critic's choice of a reading must beascribed to his personal preference. The cynic naturally prefershis own competent reading to that of another, yet he open­mindedly recognizes the right of another to be just as blithelyclosed-minded as himself. Secretly he may consider other viewsto be silly or tasteless, but since he has no objective groundsfor rejecting them, he equably tolerates all interpretive viewswhich do not conflict with known facts. On a practical level it isthus sometimes difficult to distinguish such a tough-mindedcynic from his optimistic counterpart, since both of thempreS(TVe an identical tolerance to a wide variety of readings.Both represent the same abject intellectual surrender, the sameabandonment of responSibility.

In contrast to such intellectual withdrawal there persistsamong many interpreters a_continuing faith in the possibilitiesof self-critical and rational thought. Indeed, every writteninterpretation with which I am familiar is implicitly or ex­plicitly an argument that attempts to convince a reader. Theuse of quotations, for example, aims not only to illustrate aninterpretive theory but also to support it-which is to say,validate it. Validation is practiced by the most unsystematicand arbitrary of interpreters, and the principles of validationare put into practice even by those who are most scornful ofself-critical habits of mind. Furthermore, the attempt to winadherents to an interpretive theory by means of validation isgenerally an implicit attempt to convince readers that othertheories should be rejected or modified.

168

r1

I

B. The Survival of the FittestMy purpose in this chapter wiII b .

mental principles that govern the _.e t~ descr~be the funda-and lead to objectively grounded d:alI?a~lOn.of mterpretationsflicting interpretations_des i Isc.nmlI1a:I~:ms betwecn con­ties which bedevil th >. t p t.e the clfculantles and complexi-

e II1 erpretlve entcrp· A· hparts of this book my .. . nse. s II1 t. e previous

, aln1 IS to clanfy conce t da degree of method J . I . P s an encourage

. 0 oglca self-conSCIousSome novel panacea Th . . ness, not to offerbeing put into pra~t. e pnnclples of va!idation are constantly

. . Ice, very often With h· hsopll1Sl1cation and self c ·t· I· . a Ig degree of. - n lca Integnty Nativ ·t d

tlOn to knOWledge ha J b . e WI an devo-ve a ways een capabl f h·conclusions but th . e 0 reac mg valid

. ,ey must somel1mes be def d d .Incursions of skeptical th. . en e agaInst theopportunism. Of COurse ~O:IC; which s?onsor cynicism andview all the concrete :d1

IS ar more Important to keep ineVl ence relevant to .

~retive problem than to foJIow elab ~ p.artlcular inter-110n. But a sense of confid h orate pnnclples of vahda-h ence t at such principl d .

ave a certain practic I ffi . es 0 eXIst Can. . a e 1cacIOlIsness 0 falms II1 this chapter is to h h . ne 0 my practicalmisplaced. . s ow t at SllCh confidence is not

B. THE SURVIVAL OF THE FITTEST

Although the use of quotations . .validation it is not f IS a universal technique of

, ,0 course an ad t .On the contrary, the eircuJ~rit o;~ua ~ techl1lq~e by itself.makes quotation alone a tot II y. d he mterpretlve processtion. Quotatioll is the first ~ ~ .ll1a equate means of valida-. , pnmltlve stage of thmg to demonstrate merel that _ . .e process, serv-pothcsis is legitimate and c~uld th:r~altlcular mterpretive hy­has the more ambitious go I f h .ore be correct. Validationpretation is legitimate but ath

0t s. o~~ng. not only that an inter­

is greater than or equal to th ~ l:S 1 ehhood of being correctabout the text The aim of al.do .any. other known hypothesis. . va I a1Jon IS to g. b· .

tlon to a particular int . h .Ive 0 Jcctlve sanc-erpretIve ypothesls and thereby to

169

Chapter 5: Problems and Principles of Validation

provide the only possible foundation for a consensus omniumwith regard to the text. That consensus would not, of course,endorse any particular written interpretation, but rather thewhole meaning to which several interpretations might refer--aparticular intrinsic genre capable of governing implications,rather than a particular selection of implications. Such selec­tions always vary and can do so without changing in any respectthe whole, generic meaning of the text.

The exigencies of validation should not be confused withthe exigencies of understanding. It is perfectly true that thecomplex process of construing a text always involves inter­pretive guesses as well as the testing of those guesses againstthe text and against any relevant information the interpretermight know. Thus the very process of construing a text involvesvalidations of a sort. But the process and psychology of under­standing are not reducible to a systematic structure (despite themany attempts to do so), because there is no way of compellinga right guess by means of rules and principles. Every interpreta­tion begins and ends as a guess, and no one has ever deviseda method for making intelligent guesses. The systematic siue ofinterpretation begins where the process of understandinr; ends.Understanding achieves a construction of meaning; the job ofvalidation is to evaluate the disparate constructions whichunderstanding has brought forward. Validation is thereforethe fundamental task of -interpretation as a discipline, sincewherever agreement already exists there is little practical needfor validation.

Such a consensus m.ay, of course, be quite temporary, sincethe wit of man is always devising new guesses, and his curiosityis always discovering new relevant information. A validation isachieved only with respect to known hypotheses and knownfacts: as soon as new relevant facts and/or guesses appear, theold conclusions may have to be abandoned in favor of newones. In order to avoid giving the false impression that there isanything permanent about an interpretive validation or theconsensus it aims to achieve, I now prefer the term "validation"

I

l

i..

1·

·.···.:

.

;f.1

to th B. The Survival of the Fittest. e more definitive-soundin 'd'"IS to show that a concl s' . g~ WOI verIfication." To verify

. u Ion IS true' to v l'd .concluston is probabl t ,a 1 ate IS to show that aF y rue on the b '. . f h .

rOm the nature of th h aSIS a w at IS known 2. . e case t e goal f' .

clplme must be the d' 0 Interpretation as a dis-d fi mo est one of h' .

e ned. But it also fall f ac IeVmg validations so. ows rom thmterpretation is imp!" '11 e ~ature of the case tbatconclusions based ICI y a progreSSIve discipline. Its newh ' On greater knowl d

t an the previous conelu . . h e ge, are more probableW · SIOns It as . d

Ith respect to the d' . r . reJecte .stration that a readI'n . ISCIPl.lD~ of mterpretation, the demon­

g IS va ld lmpl"e thmore than individu I . I S, erefore, a great deal. a mterpreters g II .

tlOn has to show not nIh enera Y prOVIde, A valida1ere y t at· . -but that it is the most pI 'bI an mterpretatlOn is plausible

d aUSI e One av '1 bl . . 'an boredom too imm' al a e. LIfe IS too short1 Inent to dem d hay out all the consideration h' h an t at every interpreter

but h 0 • s W IC have led t h d "w en Interpretive d' 0 suc a eClSlond . Isagrcements do '

e ge IS possible only Of Occur, genuine knowI-° d' 1 someone takes thJU Icating the issue in the Ii ht f e ~esponsibj1ity of ad-such adjudiCations.' g 0 all that IS known. That few

eXIst merely argue t -shollld be llnderte>kc A . s S 1 ongly that many more

, <. "n ..•n Interpret, ..self If he bplicvcs be l' '. er IS llsua]]y deceivinrr hin1-

~ las anythmg b . tt 0

task of such adJ'udication' f ·c er to do. Certainly the'. IS requently p t f' .

tlon and IS recognized h b ar o. an edltor's func-as Sue y sam d'

many of them find wa s of . e e ltors, though far toomatter. 3 Y escapIng their responsibility in this

~. In transcendental philosoph " . .~p~~es .to a piori certitude where~s "vah.~atlO.n""(Geltullgspru!ullg)en catIOn. I am assuming howe ven cahan means empirical

everyday usage a "valid" c' ,ver: mo~t readers feel as I do th .~y ~~ce!?,table reasoning, ~~~~~~~n t:nplles one that has been ~ea~~:~ven:d conClusion, on the other h. may not be certainly true. A

firmatl~n a!!? c~rtainty. For this re and,! strongly suggests direct con­~se of verIfIcatIOn" (Appendix I) . asfan have abandoned my earlier109 term. 10 aVOr of the less defi 't'

. OJ Ive-sound_. 3. Textual choices frequent! d '~n~erpr~tations depend upon texis ;1end. upon interpretations, just as

e ermme what the author wrote' .e aIm of the textual editor is t~or mtended to write and, no purely

170

171

,;

iI

~

Chapter 5: problems and Principles of Validation

To en.erge successfully from the rigors of an adju?ication,the victorious hypothesis must have been compared wIth everydisparate hypothesis severally or with hypoth~ses that hadalready emerged victorious over other competitors. Such aprocess is inevitable because th~ determin~tionsmust be madeby individual comparisons. An mte~pretatI?nstands or falls asa whole. As soon as the judge begms to pICk and choose ele­ments from several hypotheses, he simply introduces new,eclectic hypotheses, which must in turn st~n? or .fall as wholes.Belief in the possibilities of mere eclectiCism IS based. on afailure to understand that every interpretation ~ecessanly r~­fers to a whole meaning. It is possible that details of exeges~scan be brilliantly right while the tendency of the whole ISwrong, but the rightness of suc~ details merely confi.rms .th~notion that disparate interpretations can h~ve ~ome Imph~ations in common. The judge's primary f~IlctlOn IS not to reh~~brilliant details of inference but to decide on the ~ost vahprinciples for generating them. ~his principle of holIsm w~:)l1~dbe applicable even if the text at Issue were a small cruX Withm

a larger text. .Sometimes the arguments for lwo interpretive hypo.theses

are so strong and our knowledge so limited that a defilll~e de-.. s'ble The aim of validation therefore, IS notClSlon IS ImpoS I.· ,

necessarily to denominate:lfl individual ~ictor, but ~~t!ler toreach an objective conclusion about relative p~obabllItIes. Incomparing two interpretations i~ is always possible to reach afirm conclusion, but it may be simply that the two hypothesesare, on the basis of what is known, equally probable, an~ thatno definite choice can be made. One may conclude that mter­pretation A is more probable than. B, that it is less t::rob~ble, orthat neither of these conclusions IS warranted. ThiS third sortof decision is just as firm and objective as th~ other tw.o, ~ndit is just as much a decision. Thus, one function of valIdatIOn

mechanical system which ignores interpretation could ever reliably

reach such a determination.

172

C. The Logic of Validation: Principles of Probability

can be to show that two or more disparate interpretations areequally valid and thereby to spur further research, since twodisparate interpretations cannot both be correct. 4

This distinction between the present validity of an interpre­tation (which can be determined) and its ultimate correctness(which can never be) is not, however, an implicit admissiont~at correct interpretation is impossible. Correctness is pre­Cisely the goal of interpretation and may in fact be achieved,even though it can never be known to be achieved. We can havethe trllth without being certain that we have it, and, in the ab­sence of certainty, we can nevertheless have knowledge­knowledge of the probable. We can reach and agree upon themost probable conclusions in the light of what is known. Theobjectivity of such knowledge about texts has been and willcontinue to be disputed so long as criticism is marred by itspredilection for advocacy without any corresponding predilec­tion for adjudication, but such knowledge is nevertheless ob­jective and founded on well-established principles. The natureof those principles will be the subject of the following twosections.

C. THE LOGIC OF VALIDATION: PRINCIPLES OF

PROBABILITY

It is a distinct misfortune that influential writers on probabilitytheory should have been so predominantly oriented to mathe­matics and the natural sciences, for the logic of uncertainty

\

4. I remi~d the reader that by the term "disparate interpretation"I refer to dIfferent constructioIls (i.e. different understandinas) notmerely different interpretations. In defending the objectivit; ~f anadjudication despite the fact that it can be superseded in the future Iam following the conception of 1. M. Keynes to whom I am much in­?eb~ed. Keynes pointed out (in A Treatise on Probability) that the re­JectJOn of a pl"obability jUdgment in the light of new evidence does notchange ~he objectivit.y or the vali~ity of the earlier judgment. Its validitywas entIrely a functIOn of the eVIdence on which it was predicated.

173

Chapter 5: Problems and Principles of Validation

is fundamental to all the humane sciences as well. It is a furthermisfortune that probability theory in the eyes of the uninitiatedis a game whose rules are entirely arithmetical and statistical.But the majority of probability judgments that we draw ineveryday life are not reducible to definite numerical quantities.We are content to judge that an event is probable, highly prob­able, or almost certain, without allotting any numerical valuesto these judgments. On the basis of this observation, J. M.Keynes concluded that probabilities can be qualitative ratherthan quantitative. 5 His notion has been vigorously and justlyattacked, but even if it is true that probability judgments areat bottom quantitative, it is also true that the quantities in­volved may be vague concepts like "more," "less," "very," and"slightly."o This lack of numerical precision in no way impairsthe truth of such judgments. A man can easily and correctlyjudge that one pile of sand is larger than another without beingable to estimate the precise number of grains in each pile oreven the relative proportion of one pile to another. Further­more, under some circumstances, there might not exist anypossibility of making his judgment numerically morc precise.That is often the case both in ordinary life and in the humanesciences. It is a fallacy to equate the numerical precision ora probability judgment with its correctness. Indeed, undersome conditions, "more".and "less" are the most descriptiveand accurate judgments that can be drawn.

Since probability judgments are the staple of the historicalsciences and underlie the activity of interpretation at everypoint, from the construing of a text to the validation of aparticular construction, it is of some use in an essay on funda­mental principles to describe briefly the general foundation ofprobability judgments as they apply to interpretation. The

5. J. M. Keynes, A Treatise on Probability (Torchback ed. NewYork, 1962), pp. 34-37.

6. Even these vague concepts involve estimates of relative fre­quency. See the convincing arguments of Hans Reichenbach, Experienceand Prediction (Chicago, 1938), pp. 301-404.

174

. C. The Logic of Validation: Principles of Probability

baSIC fact about any probabilit 'u . ,.refers to a reality that'. 1 Y J dgment IS ItS uncertainty. It

IS part y unknown d h' hthe case of interpretation) b an w IC may (as inF never e known with .

orgetfulness of this basic d fi . . . certamty.ability judgments has 1 d ' e mng .charactenstlc of prob-1 . c e some theonsts ast .aclOUS notion that prob bT" ray mto the £al-

relation to the unknown r:a/i:tYa~~d1ments.bear a necessarybe evaluated accordin to th: t at theIr cor~ectnessmayreality. As Keynes righ~y observ:~bse;uent"~xp~nence of thata necessary relationship onl to it ob.abllIty JUdgments bearare based.' Because th ?' t e eVIdence on which theyit follows that a probab~l .rtea~ltdY referred to is partly unknown,. I I Y JU gment may be £ 1m relation to the know "d per ect Y correct

b nevI enee yet ineone ta out the unknown r 1"t Th: c as a statcmentparadox of all probabil~; I ;r'-d IS inevi~able and consistentfact that no matter how hY ~u gments d~nves from the simpleis inaccessible to direct ar ~e may thll1k about a reality thatuntil we do experience ~;p::Ience, we ca.nnot know what it isthere would be no po' t I.. we. could dIrectly experience it,

. . m m guessll1g about it, but if wguess about It our glles~ 1. I b e merely

, d Cal' C C W"OD P b b Tnlents are informed guesses· -rh _\ A" g. ro ~l I Ity jUdg-capable of converting an in~c~es;;b~~:la~nno I1:aglcal poter:eyknown. They are a rational mea n no~n mto son:ethmgthe absence of directly . ns of reachmg conclUSIons in

expenenced certitudeFrom the fact that a probabilit "d .

sions about somethin . . y JU gment reaches conc1u-something be in theg;~~C~~SStI:le;oexper~ence (whether thatjudgment must somehow "1 e uture), It follows that the

asslml ate its unkno b"which is kno.wn This' th wn 0 Jeet to that. . IS e central purpose of b b"l'JUdgment, and everything that" a pro a 1 Ity

. goes Into the judgm t bserves thIS purpose Th. k en su -gro" I "e un nown must somehow (even if

Pll1g Y and wrongly) be assimilated to th k .there would be no ratio 1 e nown, otherwIse

. na access at all to the u keven of a tentative nature Th' n nown-not

. IS purpose and requirement in all

7" Keynes, pp. 3-9.

175

Chapter 5: Problems and Principles of Validation

probability judgments determines the fundamental axiom andassumption that must underlie all such judgments whether t~eybe made in the service of statistical sciences or everyday lIf~:namely, that all members of the same class will tend to act mthe same way. If we cannot subsume the unknown u~~er ~6mekind of known class, then we cannot make a probabilIty Judg­ment, for we have no way of assimilating the unknown to .t~eknown. The basic and necessary assumption of all probablhty

judgments is the uniformity of the ~lass. .This assumption is far from arbItrary and can be e~sl1y ~e~

fended. The idea of a class in itself entails an idea of uIllformltyat some level for we subsume different individuals under thesame class o~ly because we observe that those individuals arcthe same in some respects. The respects in which they are thesame become the defining characteristics of the cIa.ss. Classuniformity at some level is a corollary to ~~e v~ry Idea of aclass. This point has application to probabilIty Judgm~nts byvirtue of the fact that the unknown to which they refer IS neverentirely unknown. If nothing were known about it, we wo~ldhave no object at all, but simpl)' a pure blank about whichnothing and anything could be predic~te.d. These known ~lS­pects of the object permit us to place It m a class posses~mgsome of the same traits. The n"lore we know about the object,the narrower and more reliable we can make the class. Th~n,on the basis of what we know about other individuals belo~gmgto the same class, we n"lake a guess that the unkno:vn tra~ts 9

f

any such object will be the same as the correspondmg ~ra!tsofmost individuals in the class-more often than not. ThIS IS thestructure of every probability judgment. It is a frequency judg­ment hased on our past experience of other individuals that weconceive to belong to the very same class as the un~nowl1.

Before examining some of the implications of th~s stru~turefor interpretation, I should pause to take note a! the.lmme~Iate~ly relevant problem that is said to trouble all hlstoncal sC1enc~.namely, that its objects of knowledge are not re?uI~~ and UIll­form, as in the determined natural order, but mdlVidual and

176

rI

C. The Logic of Validation: Principles of Probability

unique, as befits the human realm of freedom. This distinctionh.as been o?-~ of the main grounds for asserting that the prin­Ciples of cnhcal thought in the two branches of knowledae areradically different. But insofar as this radical separation a;pliesto the necessary use of probability judgments in both branchesof knowledge, the theory of two distinct cultures does not hold.It is simply not true that the objects of knowledge in the cul­turalscien.ces are thoroughly unique. If that were so, they couldnot be objects of knowledge. Dilthey's motto Individuum estineffabile, has as its corollary, Individuum non'est intellegibile. 8

!hi.s .maxim must hold, at any rate, for the knowledge aboutmdlvlduals that we gain through probability judgments. Theunknown traits of human beings, human actions, and human:neanings are conlcpletely inaccessible unless we ll"lanage toJudge that they belong to a class in which such traits are thusand so more often than not. If we assume that the unknowntraits are radically unique, we cannot subsmne individualsunder a class and cannot make an informed guess about theirtraits.

That probability juugments inhere in all aspects of textualinterpretation is easily demonstrated. First of all, we noticethat the construction of meaning from a text embraces elementsalready construed and accepted for the moment as beingknown, and other elements acknowledged to be unknown whichare the objects of our construing. The obvious example of thisis the construing of a crux by an appeal to a known context.But the example of a crux does not represent merely a specialcase. The object of our construing is always for the nonce aquestion mark, that is, a crux, and the basis for our choiceof a particular sort of meaning is always our appeal to what weassume we already know about the text. On the basis of thatassumption, we infer that these words coming in this place in atext of this sort probably mean thus and so. On the one side wehave the context and the sequence of words; on the other we

8. The phrase goes ba.ck to a letter from Goethe to Lavater, Sep­tember 1780. Before that, It had been an untraced scholastic maxim.

177

lie~"

c. T he Logic 0/ Validation.· P rindpies a/ Probability fnarrowly, by consequence, we can define the class to whieh itbelongs~ If ~e narrow the class so that our object becomesalmost Identical with other known objects (the more of themthe better), thell We can be less and less doubtful about theremaining unknown traits of our object.

1 hav~ al:e~d~ given one familiar example of the way doubt­f.uln~ss IS dIm1Jllshcd as the class is narrowed in alluding to thelikelIhood that a Woman will live longer than a man of the~a~e. age. Such a jUdgment, though true, is very doubtful inmdividual cases; if one could narrow the class to which theman and woman respectively belong (as insurance companiestry ~o do), the~ one might completely reverse the jUdgment andd~Clde t~at thIS particular man will probably live longer than~hIs partIcular Wornan. Similarly, one might quite correctlyJudge.that a~y ~lcdieval narrative poem is likely to be allegori­cal ~smce thIS I~ true more often than not), but a particularme.dIeval narratIve poem might belong, by virtue of certaintraIts, to a class whose members are nonallegorical more oftenthan not. Anything We can do to narrow the class such as deter­mining authorship, date, tradition, and so on, will decrease thedoubtfulness of Our prolJability judgment-that is, increase itslikelihood of being true.

Three criteria arc decisive in determining the reliability ofour guess about an unknown trait-the narrowness of the class,the number of members in it, and the frequency of the traitamong those mcmbcrs. Though the copionsness of instancesmust obviously diminish as the class narrows, we neverthelessachieve increased reliability by narrowing the class. This is trueevel)- when the narrower class has merely two members, onebeing knmvn and the other being the unknown object underscrutiny. This follows from the fundamental assumption ofprobability judgments, namely, the uniformity of the class. Fora class is narrOwed and its members made more uniform byincreasing the number of class traits. When more and more ofthese trait" are identical, the· nnknown traits of our object willhave more and marc likelihood of being identical with the

II

I

178 179

9. The principal difference is that the type-trait Illodel iIllplieswholeness of the type while the notion of class does not. See Chap. 2,Sec. D, pp. 49-50.

Chapter 5: Problems and Principles of Validation

have the meanings which we judge the words to represent inthis case. We reach those meanings entirely on the basis of ourjudgment that such meanings will occur more often in an in­stance of this sort than will other meanings, and we are able tomake that inference because we have concluded that the in­stance is of this sort (i.e. class) rather than another sort. If wecould not subsume the unknown meanings under a class onthe basis of what we already know, then we could not makesuch an inference. The exigencies of probability judgmentshere have a direct kinship with the type-trait judgments whichI described earlier in this essay. The kinship is not accidental.The type-trait model which is required to determine implica­tions is a special application of the class-instance structure inall judgments of probability.9 .. .

There is another point of identity between probabIlIty Judg­ments in general and the particular variety of them which weuse in understanding a text. I pointed out that in order to deter­mine the meaning of a word sequence it is necessary to narrowthe supposed genre of the text to such a degree that the mean­ings are no longer doubtful. I called this very narrow andparticularized conception of the text as a whole its (posi~ed)

intrinsic genre. Now this process of narrowing the genre IS aversion of the principle, well known in probability theory, ofnarrowing the class. The pdnciple arises because there are two

. questions at issue in any probability jUdg~ent: first, what,probably, are the unknown traits of the object, and second,how probable is it that our judlSllent is true? This double que~­

tion is always at issue, and our answer to the second part of Itdetermines whether we say that our conclusion itself is prob­able, highly probable, or almost certain. The degree of reliancethat we place on a probability judgmen~depends on t~is s~con­

dary decision about its likelihood of bemg true. The likelihoodwill increase the more we know about our object and the more

Chapter 5: Problems and Principles of Validation

known traits of the subclass. When we narrow the class, wedecrease the instances, but at the same time we increase thedefining traits of the class, and that is the chief goal. Thisprocess of narrowing the class is the decisive element in validat­ing interpretations, as I shall show in the next section.

D. THE LOGIC OF VALIDATION: INTERPRETIVE EVIDENCE

An interpretive hypothesis is ultimately a probability judgmentthat is supported by evidence. Normally it is compounded ofnumerous subhypotheses (i.e. constructions of individual wordsand phrases) which are also probability judgments suppo:te?by evidence. Hence, the objectivity of interpretation as a dISCI­pline depends upon our being able to make .a.n ~bjectively

grounded choice between two disparate probabIlIty Judgmentson the basis of the common evidence which supports them.Unless firm principles exist which permit such comparativejudgments to be drawn, neither interpretation nor any otherdiscipline built upon probability judgments can aspire to ob­jective knowledge. The existence of such principles does notguarantee that men will apply them-any more than the ex­istence of logic can guarantee that men shall think logically­but their existence does guarantee the possibility of objectiveknowledge, and that IS the major thesis which this book under­takes to defend.

Since we can never prove a theory to be true simply --byaccumulating favorable evidence, the only certain method ofchoosing between two hypotheses is to prove that one of themis false. In the predictive sciences this can be accomplished bydevising an experiment which conforms to the followin~ con­ditions: if theory A is true, then the result of the experImentmust be thus and so. If the result turns out not to be thus andso, theory A in its original form is permanently falsified. TheoryB on the other hand, is still consistent with the new results andm'ust be accepted for the nonce. After such a decisive experi-

180

D. The Logic of Validation: Interpretive Evidence

ment, it is stilI not certain that theory B is true, but it is certainthat theory A is false, and that is a great step forward. In thehistorical sciences such a result can seldom be achieved be-­cause decisive, falsifying data cannot be generated at will, andif such data had already been known, the two hypotheses wouldnot have been in serious competition. Sometimes, of Coursedecisive data does by good fortune turn up, but usually neithe;competing hypothesis can be falsified, and both continue aftertheir separate fashions to account for the evidence. In that casesince the direct path of falsification is closed, we have to mak~our way through a thicket of probability judgments on the basisof the evidence that We have.

As every interpreter knows, this evidence is usually con­flicting. If that were not so, we would not usually be troubledwith conflicting hypotheses. Indeed, as I observed earlier, someof the evidence Supporting one hypothesis cannot even existunder the other, since some of the "internal evidence" can begenerated only by a particular interpretation. Such incom­~ensurable, dependent evidence cannot of course serve anydlfect function in comparing interpretations, and I shall there­fore discuss later the ways in which this JJ3.nclicap can be over-­Come. But that is not in any case so crucial an issue as theproblem of directly conflicting evidence. Normally the inter­preter is faced with the dilemma that some independent evi­dence favors one hypothesis, while other independent evi­dence favors its rival. This is the normal state of affairs ininterpretation.

I bring forward a rather detailed example of such a conflicti~ Appendix I, where I quote from two disparate interpreta­hons of Wordsworth's "A Slumber Did My Spirit Seal." Atissue is the fact that the cvidence for a pessimistic and uncon­soled tone conflicts with the evidence for a tone of invincibleaffirmation. In a brief space I have tried to show that one kindof evidence outweighs the other, though my comparison (firstpublished several years ago) is not nearly so detailed as itwould have to be in order to carry universal conviction. A

181

Chapter 5: Problems and Principles of Validation

really thorough examination, bringing forward evidence whichI did not consider, might reverse the verdict or indicate thatthe evidence does not warrant a clear choice. I do not considermy little illustration to be a thoroughgoing adjudication, butit does illustrate the way interpretive evidence can be andusually is in conflict whenever interpretations are in conflict.

Another example is the conflicting evidence that supportstwo disparate modes of interpreting Blake's Songs of .Inno­cence and of Experience. Again, I have (elsewhere) laId outthe conflicting evidence, and in this case I was able to be mo~ethorough than in the case of Wordsworth's poem,10 but I stIllcannot claim that my effort is a model of adjudication, sincethe issue is still in the stage of advocacy. In order to reach areally firm decision between these two hypotheses about Bla~e,it would be wise to wait for the opposing advocates to bnngforward unfavorable evidence which I might have missed. Atthat point a more reliable adjudication could be made, sincethe advocates would then presumably have brought forwardnearly all the important relevant evidence. For an exemplarydiscussion of typically conflicting evidence, the reader may

'''p. cr f L ·d "11wish to consult M. H. Abrams Ive ypes 0 yCI as.Such examples remind us that conflicting evidence is the mainproblem in making an adjudication. ..

Thus the crucial problem in judging between dIsparate mter­pretations is usually the cQmparative weighing of relevant evi­dence. We must be able to conclude that the evidence favoring

.. one hypothesis outweighs the conflicting evidence favoring itsrival· otherwise we would have no basis for choosing onehyp;thesis over the other. Furthermore, 0.ur ~udgment abo~tthe relative weight of evidence must be objectively founded Ifwe are to claim objectivity for our decision. However, theobjectivity of our decision cannot consist (as in the convenient

10. Innocence and Experience: An Introduction to Blake (NewHaven, 1964).

11. In C. A. Patrides, ed., "Lycidas": The Tradition and the Poem(New York, 1961).

182

D. The Logic of Validation: Interpretive Evidence

device of falsification) in finding some means of avoiding adirect judicial comparison. Our decision is publicly compellingonly when our probability judgments are sanctioned by ob­jectively defined and generally accepted principles. We needprinciples for determining the admissibility (i.e. relevance) ofevidence and the relative weight of evidence.

Of course, an interpretive hypothesis need not explain allthe evidence that comes along the stream of experience. It maybe true that the best hypothesis always explains the most evi­dence, but that evidence must also be the most relevant evi­dence. Indeed, a less probable hypothesis may sometimes bebased on a greater absolute quantity of data than the moreprobable one. For example, the predictive hypothesis that thiswoman will live longer than this man is based on an immenseaccumulation of evidence embracing millions of instances buton the other hand, the evidence about the relative lif~ ex~pectancies of healthy men compared to women of the same agehaving chronic nephritis may be extremely modest-say, ahundred instances. Yet this modest sample may provide evi­dence that is much more relevant to our actual case than themillions of instances which support the contrary hypothesis.Vvc know very well that one datum does not necessarily beara significant relationship to another datum, and fortunately wedo not have to enter the Alice-in-Wonderland world of "mate­rial implication" in which we are compelled to reason: "if NewYork is a large city, then grass is green." The evidence that weare concerned with in comparing the probability of one hy­pothesis with that of another is relevant evidence, and ourimmediate concern must now be to define relevance as appliedto interpretive evidence.

Since an interpretive hypothesis is always a probabilityjudgment, it follows that the evidence which is relevant to thatjudgment must have some function in affecting the probabilitiesinvolved. If a fact or observation has no effect upon theseprobabilities, then, obviously, it is irrelevant to that particularprobability judgment. Now a probability judgment is always

183

Chapter 5: Problems and Principles of Validation

a guess about the unknown traits of a partly known instance.That guess is made on the strength of the known traits pos­sessed by other instances which belong to the same class as theinstance under scrutiny. We infer that an eighteenth-centurywriter using the w,ord "wit" probably means something generallike "intelligent competence" rather than just "clever repartee,"because the former is what other eighteenth-century writersmean by "wit" more often than not. In this case our subsumingclass is "uses of the word 'wit' in the eighteenth century," andour guess about the meaning of this instance is based on thefrequent occurrence of that meaning in other known instancesof the class. If we did not know that our text belonged to theeighteenth century, we could not subsume our instance underthat class. It follows from this structure of all probability judg­ments that evidence will be relevant which helps define thesubsuming class and which increases the number of instanceswithin the subsuming class. These two criteria of relevancebear directly on the problem of weighing evidence.

In order to decide whether a guess about a trait is probablycorrect, we necd answer only one question: does the traitoccur in the subsuming class more often than not? Quite ob­viously some guesses will be far more reliable than others: thatis to say, the probability that the probability judgment wIll becorrect varies a great -deal-.If, for example, we had fifty in­stances of the word "wit" in the eighteenth century and foundthat thirty-five of them used the word in its broad sense, the~we would, in the absence of other, narrowing data, be obligedto guess that the instance under scrutiny also conveys thatbroad sense. But while our judgment, on the basis of theknown data, would be valid, we could place very little relianceon it and would undoubtedly seek to make our guess morereliable. If, on the other hand, all known instances of "wit"in the eighteenth century conveyed the broad sense, then wecould place far more reliance on our guess, since its prob­ability of being correct would have greatly increased. Alterna­tively, if only twenty-seven of our fifty instances used the

184

D. The Logic of Validation: Interpretive Evidence

broad sense, we would be wise to conclude that the reliabilityof our guess is so small that the probabilities of the conflictingguesses are about equal and no decision is warranted in theabsence of other data. We are forced to conclude that broadsubsuming classes like "uses of 'wit' in the eighteenth century"cry out for more particular data when we want to make ourguess reliable or weighty.

The supplementary data we need are not simply more in­stances of "wit" in the eighteenth century; presumably, wealready have all the instances available. The kind of evidencewe need is information concerning those instances which aremore and more like the instance about which we are guessing.If, for example, we ascertain that our text is by a man namedRivers, and if we discover that Rivers apparently always uses"wit" to mean "clever repartee," then, on this further evidence,we would be right to guess that the present use also means"clever repartee," even though this guess is in conflict withthe guess made on the basis of all known uses of the wordin the eighteenth century. For this new, more deliniited evi­dence is far more relevant La our hypothesis than th(,~ previolls,general evidence. It serves to define a lnuch narrowe;" sub­suming class of instances, and a judgment based on this nar­rower class is necessarily more weighty and reliable as a prob­ability judgment than one based on a broader class. Thisnecessity follows, as I observed in the previous section, fromthe basic assumption of probability judgments, namely, the uni­formity of the class. By narrowing the class, we have, in effect,created a_new class far more relevant to our guess than theprevious one, and this narrower subsuming class always hasthe power to overturn (or to confirm) the evidence and theguess derived rrom the broader class. The previous frequenciesare then no longer functional. The main thing that counts atthat point is the relative frequency of our guessed-at trait with­in the new, narrower class. Here, then, is one principle forweighing conflicting interpretive evidence: the evidence of thenarrower class is always the more weighty-no matter what

185

Chapter 5: Problems and Principles oj Validation

the frequencies are within that class or any broader one thatcomprises it. The further addition of instances to our narrowclass does increase the weight and reliability of our evidence,as does an increase of the relative frequency within it, but forany given accumulation of data the evidence of the narrbwestsubsuming class is always the weightiest evidence.

This inference was already implicit in comparing the lifeexpectancy of a healthy man with that of a woman of the sameage who had chronic nephritis, but that example is remotefrom interpretive problems and is, in any case, misleading inone respect. We do not make such a judgment simply becausewe happen to know that there is a direct causal connectionbetween one trait (chronic nephritis) and another trait (nearnessof death). Relevance of evidence is not always dependent onour knowing the connection between one trait and another.It is dependent simply on our past observation that one traitwithin the subsuming class will go with another trait moreoften than not.

In the domain of interpretation the simplest and clearestexamples of the way a narrower, more fully defincd class lendsweight to evidence may be found in the work of the textualeditor. The editor of old manuscripts always has to makeprobability judgments \'ihen choosing among (or even whenrejecting) all the variant readings of his manuscripts. His soleaim is to guess correctly the word that the author intended,and in order to make this guess he has to consider an immenseamount of evidence, including (as some editors apparentlyforget) evidence about the most valid interpretation of thepassage as a whole. Most conscientious editors recognize thatno rules of thumb can lead mechanically to the most probablereading. The genealogy of the manuscripts (if known) some­times lends weight to a particular variant, but the editor knowsthat the reliability of any favored manuscript is uneven andthat its general probability of being right can be reversed byother evidence-since mistakes of transcription occur even inauthors' holographs.

186

r'~,

~.

f}:"

rk['

rI.

II!

D. The Logic oj Validation: rnterpretive Evidence

A very telling example of the way textual evidence becomesmore weighty as the class of instances is narrowed was givento me by the editor of a medieval English homilist. At onePOiI~t in the tex~ the medieval author had given the pagan god:JupIter two attnbutes. One of them, according to all the manu­scripts, was pejorative, and the second, according to many ofthem, positive. About this second attribute there was manu- ,script disagreement as to whether the word should be prymlic !

(magnificent, splendid, etc.) or pwyrlic (perverse, contrary, retc.). Of course, I cannot hope to layout all the conflicting i

evidence favoring one or the other of th~se readings, but I can rfor the purposes of illustration describe a few crucial pieces I

of evidence. First, it is in general very likely that a medieval rhomilist would be hostile to the pagan gods. Second, it is usual Ithat a homilist would not confuse matters by making his judg- ',•...•....ments only halfheartedly pejorative. Thirdly, the positive word •prymlic is unlikely, since the author rarely uses prymlic, where- ,as he lards his homilies with the pejorative pwyrlic. All of this tevidence converges to make J)}vyrlic the more probable read­ing, and if it were all the evidence we had, )ywyrlic \vould haveto be chosen. But a fourth and single piece of evidence over-­turns all these mutually supporting class frequencies: a fewlines earlier the author has written of another pagan god,Saturn, and the manuscripts show beyond reasonable doubtthat he gave Saturn two attributes, one pejorative, the otherfavorable. This second, solitary instance is similar to the cruxin so many respects-author, context, subject matter, point intime, etc.-_-that it serves to define a very narrow class underwhich the problematic reading can be sUbsumed. Obviously,this broad array of identical traits constitutes a class which isfar closer to the unknown instance than the broader and moredistant class frequencies supporting the double pejorative. Thisnarrow class is not highly reliable, since it consists of onlytwo members. Thus the choice of prymlic, while valid, is stillsomewhat doubtful. However, it is the valid choice, since ajudgment based on a narrower class is always capable of re-

187

Chapter 5: Problems and Principles of Validation

versing a judgment based on a broader one.l2 The evidence ofsuch a class about a particular trait is always weightier thanthe evidence from a broader class. Any editor with commonsense would, on the basis of the evidence given, chooseprymlic.(Informed common sense always follows the logic of prob­ability judgments, since that logic is the foundation of com­mon sense.)

This example illustrates how evidence can be weighed ac­cording to the narrowness of the subsuming class and, as acorollary, how the task of narrowing the class entails the fer­reting out of as much detailed information as possible. Evi­dence from other works of the same general period is lessweighty or reliable than evidence from other works by the sameauthor; evidence from all his works is less weighty than evi­dence from his works similar to the one at hand; evidencefrom all similar works by the author is less weighty than evi­dence taken from his similar works composed at the sameperiod as the text under scrutiny, and so on, mutatis mutandis,for other class-defining traits. Obviously, if there are no ex­ceptions, if a trait always occurs even in the broader class,theD it will always occur in the narrower class as well. Butwhen there is inconsistency in the trait's occurrence, and when,therefore, there is conflicting evidence, a decision can bereached whenever one conclusion is based on a subsumingclass that includes not only all the defining traits of the classsupporting the rival conclusion but also further defining traitsof itsown.--

The resolution of conflicting evidence in interpretation isoften less neat than this because sometimes there are piecesof conflicting evidence whose classes are incommensurable.For example, in the case of prymlic vs. }nvyrlic, we might havebeen faced with the disconcerting fact that the majority of the

12. This assumes, of course, that no other kinds of favorable or un­favorable evidence exist. (That is why I chose a simplified example.)The problem of coordinating and weighing different kinds of evidenceis discussed below.

188

..~~,

D. The Logic of Validation: Interpretive Evidenc

manuscripts give pwyrlic (though in fact they do not). If whad learned that for this text the majority reading is usuallright, our two results would conflict, and there is no obviou'way that we can compare the evidence from majority readingwith the evidence from attributes given to pagan gods in thitext-at least there is no way of comparing them on the criterion of their relative class narrowness. 13 On the other handwe might compare the reliability or weight of each judgment OJ

other grounds-for example, by showing that the majority athe manuscripts are correct only about seven times out of tenwhereas the author, when he uses similar examples to mak,his points, always treats them similarly. If we could not mabsuch a decision about reliability, we would have to conclud,that the two readings were equally probable-a situatiolwhich an honest editor acknowledges in his apparatus.

The comparing of such disparate classes immediately raise:the question of mutually incommensurable internal evidenceIt is sometimes possible to compare two conflicting interpreta :tions on the basis of internal evidence alone, but this cipportunity arises far less often than numy jnternreten; beiieveI mention skeptically some possible criteria £;)r making sudcomparisons in AppendiX I-namely, legitimacy, generic ap'propriateness, correspondence, and coherence. I observe thai'comparisons on the basis of coherence cannot be conductedsimply on the basis of internal evidence, since coherence is avariable concept. The same objection can be made against thecriterion of generic appropriateness, since the genre of thetext is als_o a variable concept-a construction or hypothesis.rather than a given. Legitimacy (i.e. the possibility that a word I

could mean what it is construed to mean) is often equallyindefinite, since legitimacy cannot be determined by fiat, but

,13. One basis for choice in editing is that of the more "difficult" I

reading. This criterion, too, is based on a class subsumption: copyistswill not usually convert an expected word to an unexpected one. Ob­viously this is only one sort of criterion which has different weight indifferent circumstances and may always be overturned if contraryevidence is weightier.

189

Chapter 5: Problems and Principles of Validation

only by observing whether contemporary readers could con­strue the word in that way. Whenever expert readers have soconstrued the word, legitimacy ceases to be a discriminatingcriterion. In short, it is usually the case that internal evidencecan discriminate between hypotheses only on the criterion ofcorrespondence. 14 That is to say, internal evidence by' itselfmight possibly indicate that one hypothesis makes functionalmore elements of the mute text than a rival hypothesis, andthe hypothesis which makes functional the greater number oftraits must, in relation to that limited evidence, be judged themore probable hypothesis.

This conclusion follows from the general probability thatstyle and sense, word choice and intended meaning, will sup­port each other. We know that the verbal choices men makehave a function in conveying their meaning more often thannot. However, it would be a grave mistake to conclude that thecorrespondence of an interpretation with the greater numberof internal traits is necessarily decisive. In the first place, thenotion that the intended meaning is the one which makes themost elements functional is not a universal law, but simply agen~ral probability whose weight varies from one kind of textto another and, indeed, from one text to another. In the secondplace, it is usually impossible, when comparing serious con­tenders, to reach a really firm conclusion on this issue, sinceone hypothesis will make-functional different traits from theother. For instance, the theory that the Wordsworth poemexpresses inconsolable grief makes highly functional the nega-

tives in

No motion has she now, no force,She neither hears nor sees.

The opposing theory that the poem is ultimately affirmativemust explain these repeated negatives as mere contrasts with

14. By "correspondence" I mean here the capacity of the interpretedsense to explain or "correspond to" the vocabulary, style, and syntaxof the text. This criterion is further discussed below.

190

D. The Logic of Validation: Interpretive Evidence

the living girl which are not so absolutely negative as the repeti­tions might indicate. On the other hand, the theory of affirma­tion makes highly functional the series, "rocks, and stones, andtrees," in the last line. The affirmative interpretation can ex­plain why living "trees" should conclude the series, whereasthe theory of inconsolability must regard "trees" only as static,inert, and passive objects like the body of the dead girl. Con­sequently, on one theory "trees" must be explained away, justas on the other theory the negatives must be. Clearly it wouldnot be warranted to conclude that one theory makes functionala greater number of textual traits than the other, for eachmakes functional different traits. That is the usual patternwhen internal evidence is compared on this quantitativecriterion.

To discover an example where the criterion of correspon­dence can lead to a clear choice, we will ordinarily have tolook beyond the disparate interpretations of experts, for if aclear choice could be made on these fairly obvious internalgrounds, then most experts will have made it before committingthemselves to print. One might expect the criterion to bedecisive in comparing, say, my students' opinion about Donue's"Valediction Forbidding Mourning," discussed earlier, withthe expert opinion that the poem is not spoken by a dyingman.15 As I observed, the students' reading is plausible, co­herent, and also legitimate, for there is not a single word in thepoem which could not legitimately be understood in Donne'sown time as our students understand it: "Mourning" couldmean grieviI!g for someone dead; to "go" could mean to die.Moreover, the idea that the souls of the speaker and his be­loved continue to live is perfectly consistent with the idea ofphysical death, while the famous simile of the compass withwhich the poem ends could reasonably be understood as sug­gesting a reunion in Heaven.

But this final simile at iast begins to diminish the explanatory

15. Seepp.73-74.

· '"';

---------------------------------------- ------.

Chapter 5: Problems and Principles of Validation

power of the students' hypothesis. Donne explicitly calls thesoul of his beloved the "fixt foot" of the compass and goes onto say:

And though it in the center sit,Yet when the other far doth rome,

It leaves, and hearkens after it,And grows erect, as that comes home.

Such wilt thou be to mee, who mustLike th'other foot, obliquely mnne;

Thy firmness makes my circle just,And makes me end, where I begunne.

The standard reading makes functional many more traits ofthis final simile than does the students' reading. It explains, forexample, why the fixed foot never has to move in order thatthere be a r~union; if the departure of the other foot is under­stood to be death, it would follow that the fixed foot would alsohave to depart in order to achieve a reunion. Furthermore, thestandard reading reveals a connection between the fixity ofthe girl (i.e. her faithfulness) and the return of the speaker.Under the students' reading most clements in the simile are notfunctional, and the simile seems loose and inept. It is quitewarranted to say, therefore, that one hypothesis makes func­tional more traitsof th~-t:e'xt than the other and is, on the basisof internal evidence alone, the more probable hypothesis.

However, our example worked neatly only because -ourstudent~ were straw men and their reading a sitting duck. Thiskind of demonstration cannot suffice to validate a single expertreading I know of at the expense of its expert rivals. Not only isit usually difficult to decide that one hypothesis makes func­tional more textual traits than another, but it is also totallyunsatisfactory to leave the matter at that. A validation requiresa consideration of all the known relevant data. For example, ifDonne had written several poems called "Valediction" and if~~ey were all spoken by dying men, that evidence would make

192

D. The Logic of Validation: Interpretive Evidence

us far less certain of our conclusion in the above case. It sohappens, of course, that Donne's other valedictions are notspoken by dying men but, rather, play on the similarities be­tween death and momentary physical absence. That furtherevidence, as it happens, supports our conclusion reached onthe criterion of correspondence. However, in making a valida­tion we cannot rest content with the fact that one single kind ofevidence favors one of the hypotheses. We want to know howthe hypotheses stand with respect to all the relevant evidencethat has been brought forward. Internal evidence is, as I havejust indicated, the evidence that is least likely to enable a deci­sion on its own grounds. Even in the case of anonymous textsof uncertain date, there always exists relevant evidence beyondsuch internal evidence, and failure to use it simply makes ourguesses unreliable and all attempts at adjudication well-nighimpossible.

Since the very limited and doubtful criterion of corres­pondence is the only one that applies to internal evidence takenby itself, we need to discover and generate other sorts ofevidence that will serve to discriminate between disparate inter­pretations. To make such discrirninations interpreters haverecourse to judgments at two distinct levels of comprehensive­ness. At the most comprehensive level, they can decide whichof two contenders is more likely to be right in its controllingor generic conception of the text. On this level, for example, wecan judge that Donne's "Valediction Forbidding Mourning"is more likely to refer to the lover's temporary physical absencethan to his death, and we make this judgment partly becausewe are famIliar with a class of poems which Donne calls "vale­dictions." I made a similar sort of generic probability judgmentabout "A Slumber Did My Spirit Seal" when I observed that,in the rest of Wordsworth's poetry written at the same period,the connection between the death of a person and the proc~sses

of nature ("earth's diurnal course") almost always implies anaffirmation of continuing life, a spark that does not die. Weknow that this may not be true in this instance, but we must

193

Chapter 5: Problems and Principles of Validation

accept the fact that such evidence does favor one hypothesisat the generic level.

But such general, large-scale probability judgments are notdecisive because there always remains a great deal of small­scale evidence which can support or overturn such a conclu­sion. This small-scale evidence is sometimes called "internal"since it comprises individual words and phrases of the text,but the appellation is misleading, since "outside" informationmust necessarily be applied in order to make a probability judg­ment about these elements in the text. For the subunits aremade to function as evidence in the following way: we positwhat the unit would have to mean under one interpretation andwhat it would have to mean under the other. Then, in isolationfrom other parts of the text, we ask which of these two sub­ordinate constructions is more likely to be correct. This carefulisolation is necessary in order to exclude arguments appealingto the coherence of a subhypothesis with the rest of the text.Such appeals to coherence are useless because, as I havepointed out, they are circular. 1G Each small-scale construc­tion will automatically be coherent wit11 the rest of the textunder the controlling conception of the text which sponsoredthe construction in the first place. Moreover, this manner ofisolating details of construction can embrace every comparisonthat might be made- according to the criterion of correspon­dence. It thus also renders that criterion supererogatory.

For example, we can isolate the opening lines of Dom~e's

"Valediction" and compare my students' construction of thesimile with that of the experts. According to the students' view,the fifth and sixth lines, like the initial simile, refer to death:

As virtuous men passe mildly away,And whisper to their soules to goe,

Whilst some of their sad friends doe say,The breath goes now, and some say no:

So let us melt and make no noise,No tears-floods, nor sigh-tempests move.

16. See Chap. 4, Sec. A, as well as Appendix I, pp. 236-38.

D. The Logic of Validation: Interpretive Evidence

Now the students' is a possible (i.e. legitimate) construction:"let our parting in death be like the peaceful death of virtuousmen." That is the sort of simile a poet might conceivably use;indeed, the romantic poets are fond of similes or metaphorswhich (in W. K. Wimsatt's terms) fuse tenor and vehicle.However, it is a far less probable interpretation than thestandard one, because it does not represent the sort of similethat Donne customarily uses. Donne habitually makes thedisparity between tenor and vehicle as striking as possible-asin fact he does (on both interpretations) elsewhere in this verypoem. Obviously, this probability judgment is not based onmerely internal evidence. It is based on the evidence thatDonne's similes are of a certain character far more often thannot, and we have gleaned this evidence from as many instancesas we could find.

Similarly, the disparate interpretations of Wordsworth'spoem compel two different constructions of the line, "Sheseemed a thing that could not feel." In the disconsolate inter­pretation, the word "thing" is regarded as a deeply ironicforeshadowing of the time when the girl would beccnne athing. Indeed, the interpretation compels that construction,since it is predicated on a jolting contrast between the livinggirl and the dead girl. Unless the word "thing" is used as ironicforeshadowing, it tends to negate rather than enforce this op­position. Yet, under the more affirmative reading the word"thing" is in no way pejorative or ironic but tends to reinforcethe idea of continuing sameness in life and death. Which ofthese two cSlllstructions of "thing" is the more probable? Ifwe consider the normal usage of Wordsworth's time, we willconclude that the first is more probable. If we consider Words­worth's habitual use of the word in poems contemporary withthis one, we will conclude that the second is more probable.Quite obviously this second conclusion, based on the narrowerclass, is the valid judgment.

But neither our conclusion about the word "thing" nor ourconclusion about Donne's first simile could be decisive byitself. Each is a single small-scale judgment which has to be

194 195

"

f•r

~.~f

Chapter 5: Problems and Principles of Validation

considered along with other small-scale judgments and large­scale Ones as well. Each of our small-scale judgments concernsthe probability of a subhypothesis which has been compelledby a particular large-scale hypothesis. We judge the relativeprobabilities of disparate individual implications generated by

Idisparate conceptions of the whole. Our purpose in makingthese small-scale judgments is always primarily to determinewhich large-scale interpretation is victorious more often thannot, for each result of comparing the probabilities of two sub­hypotheses is subsequently to be regarded as a piece of evidencefavoring one or the other generic interpretations. When one ofthe larger interpretive hypotheses is victorious more often thanthe other, we say that it "explains" more evidence and is there­fore more probable. This is often what interpreters mean whenthey say that an interpretation corresponds better to the text orexplains the text better·. As I have shown, such a description isquite inaccurate and misleading; both interpretations corres­pond to the text equally well, and both serve to explain every­thing in the text. What is reaJly meant is that the explanationsor subhypotheses implied by one interpretation turn out, on thebasis of all relevant evidence, to be usually the more probableexplanations. \'/hen the verdict of these small-scale judgmentssupports a large-scale interpretation which is also more prob­able on other grounds (as was the case with Donne's poem),then we can cOllsider- our choice to be highly reliable. How­ever, when there is conflict between these two levels, we haveto decide whether the cumulative small-scale probabilities areweighty enough (by their individual weight and their con­sistency) to overturn the large-scale probability. Usually thiswill be the case, but when it is not, we may have to reach theopposite decision or conclude that both hypotheses are equallyprobable.

My reason for refusing either to defend or reject my pre­viously published opinion about Wordsworth's poem is thatmany readers privately raised points which I had not explicitlyconsidered. To review them all would in this context be digres­sive and inconclusive, but my experience does raise a highly

196

t

lI

!t

[;t':I..

!I.

:'f

D. The Logic of Validation: Interpretive Evidence

pertinent issue with regard to the adjudication of disparateinterpretations. The really crucial necessity in reaching reliableconclusions is to accumulate numerous disparate subhypoth­eses like those I have just brought forward in illustration.Precisely such subhypotheses were brought forward by someof my dissenting readers. This illustrates the principal virtue ofthe advocacy system in interpretation as in law. The advocateshave the task of bringing forward evidence favorable to theirside and unfavorable to their opponents. In doing so, they maybring to light evidence which a judge might not have thought toconsider. But without a judge all those relevant pieces ofevidence float uselessly. Advocates are needed to discover sub­hypotheses capable of sustaining decisions, as well as othersorts of evidence capable of favoring an interpretation. How­ever, unless advocates sometimes serve as judges, none of thisactivity will actually contribute to knowledge.

I can now sum up the principles governing decisions aboutthe weight and relevance of interpretive evidence. To make areliable adjudication, all relevant evidence, "internal" and"external," should be considered. The admissibility of evidenceis determ.ined by the criterion of relevance. Evidcllce lliust beaccepted as relevant whenever it helps to define a class underwhich the object of interpretation (a word or a whole text) canbe subsumed, or whenever it adds to the instances belongingto such a class. The relative weight or reliability of a judgmentbased on such evidence is determined by the re'lative narrow­ness of the class, the copiousness of instances within the class,and the relative frequency of the trait among these instances.A judgment based on a narrower class is always more weightyor reliable than one based on a broader class-no matter howmeager the narrower class may be. When we have conflictingevidence from two disparate subsuming classes, we should tryfirst to form a third, narrower class 1;:>y combining the definingtraits of the classes. When this is impossible, and the judgmentsbased on the two classes are in conflict, we must decide whichjudgment is the more probable by comparing the copiousnessof the subsuming classes and the relative frequency of the pre-

197

, ,:", I

Chapter 5: Problems and Principles of Validation

dominant trait within the classes. However, such comparisonsare often unreliable and insecure.

The application of these principles in judging between inter­pretations occurs at two levels. On the generic level, we con­sider the relative likelihood (apart from consideration, of "in­ternal" evidence) that the text will be of one sort rather thananother. This generic guess should be conducted separately, inisolation from many of the internal traits which support it,because those traits are to some extent constituted by the gen­eric guess itself. The evidence which goes into this guess is thuspartly "extcrnal"-date, authorship, milieu, and so on-butit is necessarily founded on such indubitable "internal" traitsas vocabulary, form, and title. On the other hand, we can alsomake small-scale probability judgments about the disparateconstructions of details that have been sponsored by the dis­parate generic hypotheses. The evidence which goes into thesejudgments is likewise both internal and external, as I haveshown. Usually an effort to apply these principles will resultin the conclusion that the more probable generic guess is theone often favored by subsidiary probability judgments. V/hcnthis happy result fails to occur, the tendency of the subsidiaryjudgments is usually the more reliable evidence, since it em­braces several judgments based on fairly narrow classes. If, insuch a conflict, hO\yever, these judgments fail to tend heavilyin one direction, then no clear decision is warranted. In thecourse of making any of these probability judgments, the inter­preter's chief concern is to narrow the class; that is to say, hischief concern is to find out as much as he can about his text andall matters related to it. That everyone has always known thisconclusion is another illustration of the fact that the logic ofuncertainty is the logic of common sense.

E. METHODS, CANONS, RULES, AND PRINCIPLES

The theoretical grounding of a discipline would seem to have asits ultimate object the formulation of firmly reliable methods

198

E. Methods, Canons, Rules, and Principles

which, when followed, will lead to valid results. The theory ofinterpretation, on this view, should lead to a methodology ofinterpretation. This ideal floated from time to time before thefertile mind of Schleiermacher and guided his attempts to form­ulate reliable canons of interpretation. It was taken up withgreater confidence and system by Boeckh, who used the wordMethodologie in the title of his treatise. However, canons ofinterpretation had been in existence long before Schleier­macher wrote-in the hermeneutica sacra of biblical scholars,in the methodological asides of the Pergamene and Alexandrianschools, and most fully in the long tradition of legal interpreta­tion, several of whose rules still attest their provenance inmedieval law: noscitur a sociis; ejusdem generis; reddendosingula singulis. In literature this practical tradition persistsvery powerfully in the many handbooks which provide theundergraduate with methods of interpreting literary texts bytelling him the questions he should ask and the categories heshould apply.

The most noteworthy feature of this tradition is the varietyof the interpretive rules it has brought forth. The ruJes do notalways contradict one another, but they do proliferate in themost diverse directions. Obviously, the literary scholar needsdifferent canons from the legal or biblical scholar, and evenwithin these broad domains the canons required for one sort oftext will be different from those required for another. The legalscholar is not interested in canons which determine whether atext is allegorical, but, for that matter, the literary scholar maynot, in a given_case, require such a canon either. The scholarwho confronts an interpolated text may find useful Schleier­macher's canon that "a sentence which is uninterruptedlygoverned by the same subject or predicate as the discourseitself is to be regarded as having a direct connection with it,"but if the text is not interpolated, this canon is of course quitepointless. 17 It WOUld, on the other hand, be not only pointless

17. -Hermeneutik, p. 100.

199

Chapter 5: Problems and Principles of Validation

but misleading to apply to all texts the legal canon that "theword 'and' may be read 'or' and vice versa."18 Anyone whopores over the practical rules of interpretation quickly observesthat their range of application is always limited. No one hasever brought forward a concrete and practical canon of inter­pretation which applies to all texts, and it is my firm belief thatpractical canons are not consistently applicable even to thesmall range of texts for which they were formulated.

The most considerable attempt to formulate really generalcanons universally applicable to all texts was that of Schleier­macher, but his efforts betray some very contradictory im­pulses which indicate how uncomfortable he sometimes feltwith his project of formulating a method of interpretation.With his eye constantly on the problems of interpreting theNew Testament, he found himself generating canons about"the main topic" and "the subordinate topic" that were ob­viously more specialized than he originally intended. Evensome of his most deliberately general canons, the most generalones that have ever been formulated, do not have truly uni­versal application. His first and firmest canon, for example, isas follows: "Erster Kanan. Alles was noch einer naherenBestimmung bedarf in einer gegebenen Rede, darf nur aus demdem Verfasser und seinem urspriinglichen Publikum gemein­samen Sprachgebietbestimmt werden."19 I translate it as fol­lows: "Everything in a given text which requires fuller explana­tion must be explained and determined exclusively from thelinguistic domain common to the author and his .originalpublic." This is obvious enough, since the verbal meaning ofan author can only be a meaning which his audience couldpossibly share. That sharability is implied by the phrase "com­luon linguistic domain" and the purpose of the canon to ex­clude private meanings and anachronisms. However, thatlaudable purpose fails to embrace those texts which deliber-

18. F. J. McCaffrey. Statulory Construction: A Statement of theGeneral Rules of Statutory Construction (New York, 1953), p. 52.

19. Hermeneutik, p. 90.

200

E. Methods, Canons, Rules, and Principles

ately strive to extend their application into the future-such a:;legal texts. It is true that the "determinations" (Bestimmungen)of a legal text have to be deduced first from that commonlinguistic domain of which Schleiermacher speaks, but theymust not be limited to that domain, since they are also meanLto apply to objects and situations which did not exist when thelaw was formulated and which thus could not be compre­hended or comprehensible within that original linguistic do­main.

Does it stand any better with the second canon? "ZweiterKanan. Der Sinn eines jeden Wortes an einer gegebenen Stellemuss bestimmt werden nach seinem Zusammensein mit denendie es umgeben."20 My translation is: "The meaning of anyword in a given passage must be determined according to itscoexistence with the words that surround it." This "rule" is ofcourse a description of what every interpreter has always done,whether he knew the rule or not, since in order to construe aword at all he has to construe its function, and that cannot bedone in isolation from the larger sense which the word conveysin alliance wilh the surrounding words. Undoubtedly, there­fore, this canon has real generality, but it is perfectly useless asa practical rule. It tells everyone to do what everyone hasalways done and will continue to do without the rule, but, moreimportant, it has no capacity to enforce practical decisions.Every word is always construed in connection with its neigh­bors, and when there are alternative constructions, the sensesof the surrounding words will vary accordingly. The contextis not a fixed given, but something that can be just as variableas the word at issue. Thus, one could just as well sct down asa corollary canon that "the sense of a word must determinethe senses of the surrounding words." Both elements are vari­able and codependent, as Schleiermacher himself often impliedin his doctrine of part and whole. Indeed, the real worth inSchleiermacher's epoch-making writings on hermeneutics is

20. Ibid., p. 95.

201

I.

Chapter 5: Problems and Principles of Validation

to be found not in his canons, but in his intelligent, lengthy, anddigressive qualifications of them.

It may be set down as a general rule of interpretation thatthere are no interpretive rules which are at once general andpractical. A truly general rule will fail, as in the exarpple above,to guide us in a specific case, and a practical rule-that is, aspecific and concrete one---cannot be truly general: it mayormay not lead to the valid conclusion. Much of Schleiermacher'senergy went into qualifying both his own rules and the tradi­tional ones he inherited. He wisely said, for example, "The oldrule-do not seek beyond the text when sufficient clues ofexplanation are present in it-is of very very limited applica­tion."21 Similarly, he said of canons dealing with verbal repeti­tions: "The maxim-take as much as possible as beingtautological-is just as false as-take as much as possible asbeing emphatic."22 We find precisely the same qualificationsin the traditional canons of legal interpretation. "We ought notto deviate from the common use of the language, unless wehave very strong reasons for it"; "Where a word has a fixedtechnical meaning, it is to be taken in that sense, unless thecontext or other evidence of meaning indicates a contrarylegislative intent"; "Where the same language is used in dif­ferent parts of the statute ... it is to receive the same construc­tion . . . unless the_ general meaning and intention of the actrequire a different construction"; and so on.n Every practical

21. Ibid., p. 103.22. Ibid., p. 105.23. See, for example, the various rules of construction by Vattel,

Domat, and Lieber in Theodore Sedgwick, A Treatise on the RulesWhich Govern the Interpretation and Application of Statutory andConstitutional Law (New York, 1857), pp. 266-90. Despite his formi­dable title Sedgwick is sanely skeptical with regard to his subject: "Nordo I believe it easy to prescribe any system of rules of interpretation forcases of ambiguity in written language that will really avail to guidethe mind in the decision of doubt. ... It would seem as vain to attemptto frame positive and fixed rules of interpretation as to endeavor, inthe same way, to define the mode by which the mind shall draw con­clusions from testimony" (p. 228).

202

E. Methods, Canons, Rules, and Principles

rule of interpretation has an implicit "unless" after it, whichmeans, of course, that it is not really a rule.

What then is the status of the many traditional canons andmaxims of interpretation, and what is their purpose? Clearly,they are provisional guides, or rules of thumb. In the absenceof compelling indications to the contrary we follow them be­cause they hold true more often than not. In other words, thepractical canons of interpretation are preliminary probabilityjudgments based on past experience. More often than not alegal text will mean the same thing when it uses the same words-and there are very plausible reasons why this should be so.However, since all practical interpretive canons are merelypreliminary probability judgments, two consequences followwith regard to their intelligent application. First, the canon ismore reliable the narrower its intended range of application.Practical canons that apply to a very strictly limited class oftexts will be more reliable for those texts than canons whichlay claim to broader application. Second, since any interpretivecanon can be overturned by subsuming the text under a stillnarrower class in which the canon fails to hold or holds by sucha small majority that it becomes doubtful, it follows that inter­pretive canons are often relatively useless baggage. vVhen theyare general, they cannot compel decisions, and even when theyare narrowly practical, they can be overturned. The importantpoint about a rule of thumb is that it is not a rule.

The notion that a reliable methodology of interpretation canbe built upon a set of canons is thus a mirage. Precookedmaxims carry less authority than informed probability judg­ments about particular cases, and verbal constructions cannotpossibly be governed by any methods. No possible set of rulesor rites of preparation can generate or compel an insight intowhat an author means. The act of understanding is at first agenial (or a mistaken) guess, and there are no methods formaking guesses; no rules for generating insights. The methodi­cal activity of interpr'etation commences when we begin to testand criticize our guesses. These two sides of the interpretive

203

Chapter 5: Problems and Principles of Validation

process, the hypothetical and the critical, are not of courseneatly separated when we are pondering a text, for we areconstantly testing our guesses both large and small as wegradually build up a coherent structure of meaning. We wantto be sure that we are getting the matter right, and we are con­stantly asking ourselves whether a guess is probable in the lightof what we know about the text so far. But the fact tl~at thesetwo activities require and accompany one another in the processof understanding should not lead us to confuse the whimsicallawlessness of guessing with the ultimately methodical char­acter of testing. Both processes are necessary in interpretation,but only one of them is governed by logical principles. Thelegal phrase "canons of construction" is thus a typical mis­nomer which reflects a long-standing confusion of the twoprocesses. There can be no canons of construction, but onlycanons which help us to choose between alternative meaningsthat have already been construed from the text.

Schleiermacher, despite his flirtation with canons of con­struction, stated this distinction imprecisely but vividly:

For thc whole enterprise of interpretation there arc fromthe start two functions-the divinatory and the compara­tive-which reflect back on one another and should notbe isolated from one another. The divinatory is the func­tion by whidl onE as it were transforms himself into theauthor, seeking directly what is individual. The compara­tive function regards what is to be understood first assomething general, and then finds out what is unique bycomparing it with other things subsumed under the samegeneral idea. The former is the female force in humanknowledge, the latter the male. 24

What Schleiermacher calls the "divi-natory function" is theproductive guess or hypothesis for which no rules can be

24. H CTI71ell eutik, p. 109.

204

E. Methods, Canons, Rules, and Principles

formulated but without which the process of interpretationcannot even begin. The critical, masculine function, on theother hand, cannot bring forth, but it can judge and test.Sehleiermacher calls it "comparative" partly because he hasrecognized that interpretive guesses are always tested by mak­ing comparisons, i.e. by subsuming the object of interpretationunder a class of similar instances. He thus recognized implicitlythe comparative nature of probability judgments, and thoughhe rightly insisted that the divinatory and comparative func­tions go together, he failed to notice that one function is alwaysprior to the other, that female intuition brings forth the ideaswhich the comparative male judgment then tests and eitheraccepts or rejects.

Despite his metaphorical imprecision Schleiermacher isworth quoting for another reason. He suggests that the femaledivinatory function and the male comparative function are thetwo principal forces not only in interpretation but in humanknowledge generally. The implications of that insight stretchbeyond the currently fashionable discussion of the oppositionbetween scientific and humanistic cultures and their respective"methods." What is at stake is not some ideal fusion of theseparate cultures and their modes of thought, but the rightof interpretation (and implicitly all humanistic disciplines) toclaim as its object genuine knowledge. The two forces thatSchleiermacher perceived in interpretation and in human think­ing generally are versions of two processes that are indeedcomprised in every realm of thought that can lay claim toknowledge. Thus Sir Peter Medawar states:

What are usually thought of as two alternative and indeedcompeting accounts of one process of thought are in factaccounts of the two successive and complementary epi­sodes of thought that occur in every advance of scientificunderstanding.... The chief weakness of Millian induc­tion was its failure to distinguish between the acts of mindinvolved in discovery and in proof. ... Mill thought that

205

Chapter 5: Problems and Principles of Validation

his process of "induction" could fulfill the same twofunctions; but, alas, mistakenly, for it is not the origin butonly the acceptance of hypotheses that depends upon theauthority of logic.... Obviously "having an idea" is animaginative exploit of some kind, the work of a singlemind; obviously "trying it out" must be a ruthlesslycritical process to which many skills and many hands maycontribute. 25

While there is not and cannot be any method or model ofcorrect interpretation, there can be a ruthlessly critical processof validation to which many skills and many hands may con­tribute. Just as any individual act of interpretation comprisesboth a hypothetical and a critical function, so the discipline ofinterpretation also comprises the having of ideas and the test­ing of them. At the level of the discipline these two "moments"or "episodes" can be separated in a way that they cannot bein the course of construing a text, for any written interpretationis a hypothesis implying a" number of subhypotheses, all ofwhich are open to examination. Conflicting interpretations canbe subjected to scrutiny in the light of the relevant evidence,and objective conclusions can be reached. Of course, imagina­tion is required--a divinatory talent like that needed to makeinterpretive guesses-simply to discover highly relevant evi­dence. Devisingsub~idiaryinterpretive hypotheses capable ofsponsoring probability decisions is not in principle differentfrom devising experiments which can sponsor deeisiclDs be­tween hypotheses in the natural sciences. But although thedivinatory faculty is essential even in the validating process,the essence of that process is the making of judgments on thebasis of all the relevant evidence that has so far been broughtforward, and such judgments can be made in the light of day.

Even the fact that some un-self-critical or fractious soulsmight stubbornly refuse assent to conclusions so reached does

25. "Anglo-Saxon Attitudes," Encounter, 25 (August 1965), 54.Medawar acknowledges his debt to Karl Popper.

206

E. Methods, Canons, Rules, and Principles

not exclude such conclusions from the domain of genuineknowledge. For when a scholar has said, "Here is all the rele­vant evidence that has been brought forward, and here are theconclusions which that evidence requires," his statement is nolonger subject merely to opposition by rhetorical posturing.His claim can be shown to be false-either because he hasoverlooked some of the known evidence or because he hasmade a mistake in logic. Such an exposure of his oversight orhis mistake can objectively overturn his conclusion, but nothingelse can. His conclusion must stand until new evidence isbrought forward.

The discipline of interpretation is founded, then, not on amethodology of construction but on a logic of validation. Theprinciples of that logic, outlined in the preceding sections ofthis chapter, are essentially the principles which underlie thedrawing of objective probability judgments in all domains ofthought. The inevitable tendency of those logical principles isaway from generalized maxims and toward an increasing par­ticularity of relevant observations. The proper realm for gen­eralizations in hermeneutics turns out to bc the realm of prin­ciples, not of methods, for the principles underlying probabilityjudgments require that every practical interpretive problem besolved in its particularity and not in accordance with maximsand approaches which usurp the name of theory. Nevertheless,despite its practical concreteness and variability, the root prob­lem of interpretation is always the same-to guess what theauthor meant. Even though we can never be certain that ourinterpretive guesses are correct, we know that they can becorrect and tnat· the goal of interpretation as a disciplinc isconstantly to increase the probability that they are correct. Inthe earlier chapters of this book, I showed that only one inter­pretive problem can be answcred with objectivity: "What, inall probability, did the author mean to convey?" In this finalchapter, I have tried to show more particularly wherein thatobjectivity lies. It lies in our capacity to say on fum principles,"Yes, that answer is valid" or "No, it is not."

207

, ,